refactor: remove type field from models_huggingface.yaml and include type in dest paths

- Prepended ComfyUI model type folder (checkpoints/, clip/, vae/, etc.) to all dest paths
- Removed separate 'type' field from all model entries
- Consolidated SD3.5 duplicate entries (5 → 1)
- Simplified model configuration by embedding directory structure directly in destination paths

This change eliminates the need to parse the 'type' field separately in artifact_huggingface_download.sh,
making the configuration more explicit and easier to understand.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

.env.example

@@ -15,10 +15,33 @@ HF_TOKEN=hf_your_token_here
 # Get it with: tailscale ip -4
 # GPU_TAILSCALE_IP=100.100.108.13
 
+# ============================================================================
+# CivitAI API Key (Optional, for NSFW models)
+# ============================================================================
+# Get your API key from: https://civitai.com/user/account
+# Required for downloading NSFW models with artifact_civitai_download.sh
+# CIVITAI_API_KEY=your_civitai_api_key_here
+
+# ============================================================================
+# Tailscale VPN (Optional, for remote access)
+# ============================================================================
+# Get your authkey from: https://login.tailscale.com/admin/settings/keys
+# TAILSCALE_AUTHKEY=tskey-auth-your_tailscale_authkey_here
+
+# ============================================================================
+# WebDAV Sync (Optional, for output synchronization)
+# ============================================================================
+# WebDAV credentials for syncing ComfyUI outputs to cloud storage
+# WEBDAV_URL=https://webdav.example.com/
+# WEBDAV_USERNAME=your_username
+# WEBDAV_PASSWORD=your_password
+# WEBDAV_REMOTE_PATH=/path/to/comfyui/outputs
+
 # ============================================================================
 # Notes
 # ============================================================================
 # - HF_TOKEN is the only required variable for basic operation
+# - CIVITAI_API_KEY is required for downloading NSFW models from CivitAI
 # - Models will be cached in /workspace/ directories on RunPod
 # - Orchestrator automatically manages model switching
 # - No database credentials needed (stateless architecture)

.gitea/workflows/build-docker-image.yml

@@ -0,0 +1,114 @@
+name: Build and Push RunPod Docker Image
+
+on:
+  push:
+    branches:
+      - main
+    tags:
+      - 'v*.*.*'
+  pull_request:
+    branches:
+      - main
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: 'Custom tag for the image'
+        required: false
+        default: 'manual'
+
+env:
+  REGISTRY: dev.pivoine.art
+  IMAGE_NAME: valknar/runpod-ai-orchestrator
+
+jobs:
+  build-and-push:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+        with:
+          platforms: linux/amd64
+
+      - name: Log in to Gitea Container Registry
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ gitea.actor }}
+          password: ${{ secrets.REGISTRY_TOKEN }}
+
+      - name: Extract metadata (tags, labels)
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          tags: |
+            # Tag as 'latest' for main branch
+            type=raw,value=latest,enable={{is_default_branch}}
+            # Tag with branch name
+            type=ref,event=branch
+            # Tag with PR number
+            type=ref,event=pr
+            # Tag with git tag (semver)
+            type=semver,pattern={{version}}
+            type=semver,pattern={{major}}.{{minor}}
+            type=semver,pattern={{major}}
+            # Tag with commit SHA
+            type=sha,prefix={{branch}}-
+            # Custom tag from workflow_dispatch
+            type=raw,value=${{ gitea.event.inputs.tag }},enable=${{ gitea.event_name == 'workflow_dispatch' }}
+          labels: |
+            org.opencontainers.image.title=RunPod AI Orchestrator
+            org.opencontainers.image.description=Minimal Docker template for RunPod deployment with ComfyUI + vLLM orchestration, Supervisor process management, and Tailscale VPN integration
+            org.opencontainers.image.vendor=valknar
+            org.opencontainers.image.source=https://dev.pivoine.art/${{ gitea.repository }}
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: ./Dockerfile
+          platforms: linux/amd64
+          push: ${{ gitea.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          cache-from: type=registry,ref=${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:buildcache
+          cache-to: type=registry,ref=${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:buildcache,mode=max
+
+      - name: Generate image digest
+        if: gitea.event_name != 'pull_request'
+        run: |
+          echo "### Docker Image Published :rocket:" >> $GITEA_STEP_SUMMARY
+          echo "" >> $GITEA_STEP_SUMMARY
+          echo "**Registry:** \`${{ env.REGISTRY }}\`" >> $GITEA_STEP_SUMMARY
+          echo "**Image:** \`${{ env.IMAGE_NAME }}\`" >> $GITEA_STEP_SUMMARY
+          echo "" >> $GITEA_STEP_SUMMARY
+          echo "**Tags:**" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY
+          echo "${{ steps.meta.outputs.tags }}" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY
+          echo "" >> $GITEA_STEP_SUMMARY
+          echo "**Pull command:**" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`bash" >> $GITEA_STEP_SUMMARY
+          echo "docker pull ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:latest" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY
+          echo "" >> $GITEA_STEP_SUMMARY
+          echo "**Use in RunPod template:**" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY
+          echo "Container Image: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:latest" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY
+
+      - name: PR Comment - Image built but not pushed
+        if: gitea.event_name == 'pull_request'
+        run: |
+          echo "### Docker Image Built Successfully :white_check_mark:" >> $GITEA_STEP_SUMMARY
+          echo "" >> $GITEA_STEP_SUMMARY
+          echo "Image was built successfully but **not pushed** (PR builds are not published)." >> $GITEA_STEP_SUMMARY
+          echo "" >> $GITEA_STEP_SUMMARY
+          echo "**Would be tagged as:**" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY
+          echo "${{ steps.meta.outputs.tags }}" >> $GITEA_STEP_SUMMARY
+          echo "\`\`\`" >> $GITEA_STEP_SUMMARY

CLAUDE.md

@@ -4,49 +4,39 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Overview
 
-This is a lightweight, process-based AI model orchestrator designed for RunPod GPU instances (specifically RTX 4090 with 24GB VRAM). It manages sequential loading of multiple large AI models on a single GPU, providing OpenAI-compatible API endpoints for text, image, and audio generation.
+This is a lightweight AI model serving infrastructure for RunPod GPU instances (specifically RTX 4090 with 24GB VRAM). It provides OpenAI-compatible API endpoints for text, image, and audio generation via dedicated vLLM servers and ComfyUI.
 
 **Key Design Philosophy:**
-- **Sequential model loading** - Only one model active at a time to fit within GPU memory constraints
+- **Direct vLLM access** - Dedicated FastAPI servers for each text model (no proxy layer)
-- **Process-based architecture** - Uses Python subprocess instead of Docker-in-Docker for RunPod compatibility
+- **Process-based architecture** - Managed by Supervisor for reliability
-- **Automatic model switching** - Orchestrator detects request types and switches models on-demand
+- **Simple service management** - Start/stop models independently as needed
 - **OpenAI-compatible APIs** - Works seamlessly with LiteLLM proxy and other AI tools
 
 ## Architecture
 
 ### Core Components
 
-1. **Orchestrator** (`model-orchestrator/orchestrator_subprocess.py`)
+1. **vLLM Servers** (`vllm/server_qwen.py`, `vllm/server_llama.py`)
-   - FastAPI proxy server listening on port 9000
+   - Dedicated FastAPI servers for each text model
-   - Manages model lifecycle via Python subprocesses
+   - Qwen 2.5 7B on port 8000, Llama 3.1 8B on port 8001
-   - Routes requests to appropriate model services
+   - Direct AsyncLLMEngine access for optimal streaming performance
-   - Handles sequential model loading/unloading
+   - max_model_len=20000 optimized for 24GB VRAM
 
-2. **Model Registry** (`model-orchestrator/models.yaml`)
+2. **ComfyUI** (`comfyui/`)
-   - YAML configuration defining available models
+   - Image/video/audio generation server on port 8188
-   - Specifies: type, framework, service script, port, VRAM requirements, startup time
+   - FLUX, SDXL, CogVideoX, MusicGen models
-   - Easy to extend with new models
+   - Advanced workflow system with custom nodes
 
-3. **Model Services** (`models/*/`)
+3. **Supervisor** (`supervisord.conf`)
-   - Individual Python servers running specific AI models
+   - Process manager for all AI services
-   - vLLM for text generation (Qwen 2.5 7B, Llama 3.1 8B)
+   - Auto-restart, centralized logging
-   - ComfyUI for image/video/audio generation (FLUX, SDXL, CogVideoX, MusicGen)
+   - Web interface on port 9001
 
 4. **Ansible Provisioning** (`playbook.yml`)
    - Complete infrastructure-as-code setup
    - Installs dependencies, downloads models, configures services
    - Supports selective installation via tags
 
-### Why Process-Based Instead of Docker?
-
-The subprocess implementation (`orchestrator_subprocess.py`) is preferred over the Docker version (`orchestrator.py`) because:
-- RunPod instances run in containers - Docker-in-Docker adds complexity
-- Faster model startup (direct Python process spawning)
-- Simpler debugging (single process tree)
-- Reduced overhead (no container management layer)
-
-**Note:** Always use `orchestrator_subprocess.py` for RunPod deployments.
-
 ## Common Commands
 
 ### Repository Management with Arty
@@ -57,17 +47,16 @@ This project uses Arty for repository and deployment management. See `arty.yml`
 # Clone all repositories (fresh deployment)
 arty sync --env prod          # Production: Essential nodes only
 arty sync --env dev           # Development: All nodes including optional
-arty sync --env minimal       # Minimal: Just orchestrator + ComfyUI base
+arty sync --env minimal       # Minimal: Just vLLM + ComfyUI base
 
 # Run deployment scripts
 arty run setup/full           # Show setup instructions
 arty run models/link-comfyui  # Link downloaded models to ComfyUI
 arty run deps/comfyui-nodes   # Install custom node dependencies
-arty run services/start       # Start orchestrator
+arty run services/start       # Start supervisor
 arty run services/stop        # Stop all services
 
 # Health checks
-arty run health/orchestrator  # Check orchestrator
 arty run health/comfyui      # Check ComfyUI
 arty run check/gpu           # nvidia-smi
 arty run check/models        # Show cache size
@@ -119,14 +108,16 @@ arty run services/status            # Same via arty
 supervisorctl status                # Direct supervisor command
 
 # Individual service control
-supervisorctl start orchestrator    # Start orchestrator
+supervisorctl start vllm-qwen       # Start Qwen vLLM server
+supervisorctl start vllm-llama      # Start Llama vLLM server
 supervisorctl restart comfyui       # Restart ComfyUI
-supervisorctl stop orchestrator     # Stop orchestrator
+supervisorctl stop vllm-qwen        # Stop Qwen vLLM server
 arty run services/restart-comfyui   # Restart ComfyUI via arty
 
 # View logs
 supervisorctl tail -f comfyui       # Follow ComfyUI logs
-supervisorctl tail -f orchestrator  # Follow orchestrator logs
+supervisorctl tail -f vllm-qwen     # Follow Qwen vLLM logs
+supervisorctl tail -f vllm-llama    # Follow Llama vLLM logs
 arty run services/logs              # Follow ComfyUI logs via arty
 
 # Web interface
@@ -141,20 +132,80 @@ arty run services/logs              # Follow ComfyUI logs via arty
 
 **Services managed:**
 - `comfyui` - ComfyUI server (port 8188, autostart enabled)
-- `orchestrator` - Model orchestrator (port 9000, autostart disabled)
+- `vllm-qwen` - Qwen 2.5 7B vLLM server (port 8000, autostart disabled)
+- `vllm-llama` - Llama 3.1 8B vLLM server (port 8001, autostart disabled)
+- `webdav-sync` - WebDAV sync service for ComfyUI outputs
 
+### GPU Memory Management and Mode Switching
+
+**VRAM Constraints (RTX 4090 - 24GB total):**
+
+The GPU has limited memory, which requires manual service switching:
+
+| Service | Model | VRAM Usage | Compatible With |
+|---------|-------|------------|-----------------|
+| ComfyUI | FLUX Schnell FP16 | ~23GB | None (uses all VRAM) |
+| ComfyUI | SDXL Base | ~12GB | Small vLLM models |
+| vLLM | Qwen 2.5 7B | ~14GB | None (conflicts with ComfyUI) |
+| vLLM | Llama 3.1 8B | ~17GB | None (conflicts with ComfyUI) |
+
+**Mode Switching Workflow:**
+
+Since ComfyUI and vLLM models cannot run simultaneously (they exceed 24GB combined), you must manually switch modes:
+
+**Switch to Text Generation Mode (vLLM):**
+```bash
+# 1. Stop ComfyUI
+supervisorctl stop comfyui
+
+# 2. Start desired vLLM server
+supervisorctl start vllm-qwen    # or vllm-llama
+
+# 3. Verify
+supervisorctl status
+nvidia-smi  # Check VRAM usage
+```
+
+**Switch to Image/Video/Audio Generation Mode (ComfyUI):**
+```bash
+# 1. Stop vLLM servers
+supervisorctl stop vllm-qwen vllm-llama
+
+# 2. Start ComfyUI
+supervisorctl start comfyui
+
+# 3. Verify
+supervisorctl status
+nvidia-smi  # Check VRAM usage
+```
+
+**Access via Supervisor Web UI:**
+
+You can also switch modes using the Supervisor web interface:
+- URL: `https://supervisor.ai.pivoine.art` (via VPS proxy) or `http://100.114.60.40:9001` (direct Tailscale)
+- Username: `admin`
+- Password: `runpod2024`
+- Click "Start" or "Stop" buttons for each service
+
+**Integration with LiteLLM:**
+
+The vLLM servers integrate with LiteLLM on the VPS for unified API access:
+- vLLM models (qwen-2.5-7b, llama-3.1-8b) available via direct connections
+- Requests route directly to vLLM servers (ports 8000, 8001)
+- Environment variables `GPU_VLLM_QWEN_URL` and `GPU_VLLM_LLAMA_URL` configure connections
+- LiteLLM config uses `os.environ/GPU_VLLM_*_URL` syntax for dynamic URLs
 
 ### Testing
 
 ```bash
-# Health check
+# Health check (Qwen vLLM server)
-curl http://localhost:9000/health
+curl http://localhost:8000/health
 
 # List available models
-curl http://localhost:9000/v1/models
+curl http://localhost:8000/v1/models
 
 # Test text generation (streaming)
-curl -s -N -X POST http://localhost:9000/v1/chat/completions \
+curl -s -N -X POST http://localhost:8000/v1/chat/completions \
   -H 'Content-Type: application/json' \
   -d '{
     "model": "qwen-2.5-7b",
@@ -163,12 +214,10 @@ curl -s -N -X POST http://localhost:9000/v1/chat/completions \
     "stream": true
   }'
 
-# Test image generation
+# Test ComfyUI (image generation)
-curl -X POST http://localhost:9000/v1/images/generations \
+curl -X POST http://localhost:8188/prompt \
   -H 'Content-Type: application/json' \
   -d '{
-    "model": "flux-schnell",
-    "prompt": "A serene mountain landscape at sunset",
     "size": "1024x1024"
   }'
 ```
@@ -197,65 +246,132 @@ curl -X POST http://localhost:9000/v1/images/generations \
 - `systemd` - Configure systemd services (use `never` - not for RunPod)
 - `validate` - Health checks (use `never` - run explicitly)
 
-### Adding New Models
+### Adding New vLLM Models
 
-1. **Add model definition to `model-orchestrator/models.yaml`:**
+1. **Create dedicated server script** (e.g., `vllm/server_mistral.py`):
-
-```yaml
-llama-3.1-8b:
-  type: text
-  framework: vllm
-  service_script: models/vllm/server_llama.py
-  port: 8001
-  vram_gb: 17
-  startup_time_seconds: 120
-  endpoint: /v1/chat/completions
-  description: "Llama 3.1 8B Instruct"
-```
-
-2. **Create service script** (`models/vllm/server_llama.py`):
 
 ```python
-import os
+#!/usr/bin/env python3
-from vllm.entrypoints.openai.api_server import run_server
+from vllm import AsyncLLMEngine, AsyncEngineArgs
+# ... (copy structure from server_qwen.py)
 
-model = "meta-llama/Llama-3.1-8B-Instruct"
+model_name: str = "mistralai/Mistral-7B-Instruct-v0.3"
-port = int(os.getenv("PORT", 8001))
+port = 8002  # Choose unique port
-run_server(model=model, port=port)
+```
+
+2. **Add supervisor configuration** (`supervisord.conf`):
+
+```ini
+[program:vllm-mistral]
+command=vllm/venv/bin/python vllm/server_mistral.py
+directory=.
+autostart=false
+autorestart=true
+environment=HF_HOME="../huggingface_cache",HF_TOKEN="%(ENV_HF_TOKEN)s"
 ```
 
 3. **Download model** (handled by Ansible playbook or manually via HuggingFace CLI)
 
-4. **Restart orchestrator:**
+4. **Add to LiteLLM config** (on VPS):
+
+```yaml
+- model_name: mistral-7b
+  litellm_params:
+    model: hosted_vllm/openai/mistral-7b
+    api_base: os.environ/GPU_VLLM_MISTRAL_URL
+    api_key: "EMPTY"
+```
+
+5. **Start the service:**
 
 ```bash
-bash scripts/stop-all.sh && bash scripts/start-all.sh
+supervisorctl start vllm-mistral
 ```
 
+### Downloading NSFW Models from CivitAI
+
+The project includes dedicated tooling for downloading NSFW models from CivitAI, separate from the HuggingFace downloader.
+
+**Files:**
+- Configuration: `models_civitai.yaml` - Defines available CivitAI models
+- Downloader: `artifact_civitai_download.sh` - Beautiful CLI tool for downloading models
+
+**Requirements:**
+- CivitAI API key (set in `.env` file as `CIVITAI_API_KEY`)
+- Get your key from: https://civitai.com/user/account
+
+**Available Models (Essential):**
+1. **LUSTIFY v7.0 GGWP** (6.31GB) - Photoreal NSFW checkpoint for women
+2. **Pony Diffusion V6 XL** (6.5GB) - Anime/furry NSFW with balanced content
+3. **RealVisXL V5.0** (6.8GB) - Photorealistic NSFW model
+
+**Usage:**
+
+```bash
+# Download all essential NSFW models (19.61GB)
+artifact_civitai_download.sh
+
+# Download only (no symlinks)
+artifact_civitai_download.sh download
+
+# Create symlinks only (models already downloaded)
+artifact_civitai_download.sh link
+
+# Custom configuration file
+artifact_civitai_download.sh -c /path/to/models_civitai.yaml
+
+# Help
+artifact_civitai_download.sh --help
+```
+
+**Directory Structure:**
+- Models downloaded to: `/workspace/models/civitai/` (cache)
+- Symlinks created in: `/workspace/ComfyUI/models/checkpoints/`
+
+**Features:**
+- Beautiful purple/magenta themed CLI (vs cyan for HuggingFace)
+- Automatic retry with exponential backoff for rate limiting
+- Resume interrupted downloads
+- Progress bars and status indicators
+- Parallel download support (respects rate limits)
+
+**Configuration Example (`models_civitai.yaml`):**
+
+```yaml
+model_categories:
+  nsfw_checkpoints:
+    - name: lustify-v7-ggwp
+      version_id: "1094291"
+      model_id: "573152"
+      description: "LUSTIFY v7.0 GGWP - Photoreal NSFW checkpoint"
+      size_gb: 6.31
+      essential: true
+      type: checkpoints
+```
+
+**Installation Profiles:**
+- `essential`: 3 models (LUSTIFY, Pony, RealVisXL) - 19.61GB, 30-60 min
+- `complete`: All 6 models - 39.21GB, 1-2 hours
+- `realistic_only`: LUSTIFY + RealVisXL - 13.11GB, 20-40 min
+- `anime_only`: Pony + WAI-NSFW - 13.0GB, 20-40 min
+
+**Important Notes:**
+- All models are SDXL-based (require ~12GB VRAM)
+- Cannot run simultaneously with FLUX models (both use ~23GB)
+- Can run with smaller models like standard SDXL (~12GB total)
+- Rate limiting: 5 second delay between downloads (configurable)
+
 ## Key Implementation Details
 
-### Model Switching Logic
+### Direct vLLM Server Architecture
 
-The orchestrator automatically switches models based on:
+Each vLLM model runs as a dedicated FastAPI server:
-- **Endpoint path** - `/v1/chat/completions` → text models, `/v1/images/generations` → image models
+- Independent process management via Supervisor
-- **Model name in request** - Matches against model registry
+- Direct AsyncLLMEngine access for optimal performance
-- **Sequential loading** - Stops current model before starting new one to conserve VRAM
+- OpenAI-compatible `/v1/chat/completions` endpoint
+- Streaming support with proper delta formatting
 
-See `orchestrator_subprocess.py:64-100` for process management implementation.
+See `vllm/server_qwen.py` for reference implementation.
-
-### Model Registry Structure
-
-Each model in `models.yaml` requires:
-- `type` - text, image, or audio
-- `framework` - vllm, openedai-images, audiocraft, comfyui
-- `service_script` - Relative path to Python/shell script
-- `port` - Service port (8000+)
-- `vram_gb` - GPU memory requirement
-- `startup_time_seconds` - Max health check timeout
-- `endpoint` - API endpoint path
-- `description` - Human-readable description
-
-### Environment Variables
 
 Set in `.env` file:
 - `HF_TOKEN` - **Required** - HuggingFace API token for model downloads

COMFYUI_MODELS.md

@@ -1,320 +0,0 @@
-# ComfyUI Models & Nodes - Usage Guide
-
-This document explains how to use the extended playbook.yml for installing ComfyUI models and custom nodes.
-
-## What's Been Added
-
-### Model Categories
-
-**Image Generation Models:**
-- FLUX.1 Schnell (23GB, essential) - Fast 4-step inference
-- FLUX.1 Dev (23GB, optional) - Balanced quality/speed
-- SDXL Base 1.0 (7GB, essential) - Industry standard
-- SDXL Refiner 1.0 (6GB, optional) - Enhances base output
-- SD 3.5 Large (18GB, optional) - Latest Stability AI
-
-**Video Generation Models:**
-- CogVideoX-5B (20GB, essential) - Professional text-to-video
-- SVD (8GB, essential) - 14 frame image-to-video
-- SVD-XT (8GB, optional) - 25 frame image-to-video
-
-**Audio Generation Models:**
-- MusicGen Small (3GB, optional) - Fast generation
-- MusicGen Medium (11GB, essential) - Balanced quality
-- MusicGen Large (22GB, optional) - Highest quality
-
-**Support Models:**
-- CLIP H (2GB, essential) - For SD 1.5 IP-Adapter
-- CLIP G (7GB, essential) - For SDXL IP-Adapter
-- SigLIP (2GB, essential) - For FLUX models
-
-### Custom Nodes
-
-**Essential Nodes (installed by default):**
-- ComfyUI-Manager - Install/manage custom nodes and models
-- ComfyUI-VideoHelperSuite - Video operations
-- ComfyUI-AnimateDiff-Evolved - Video generation
-- ComfyUI_IPAdapter_plus - Style transfer
-- ComfyUI-Impact-Pack - Auto face enhancement
-- comfyui-sound-lab - Audio generation
-
-**Optional Nodes:**
-- ComfyUI-CogVideoXWrapper - CogVideoX integration
-- Comfyui-Inspire-Pack - Additional tools
-- ComfyUI-Advanced-ControlNet - Advanced control
-- ComfyUI-3D-Pack - 3D asset generation
-
-## Usage Examples
-
-### Quick Setup (Essential Only)
-
-Install ComfyUI with only essential models and nodes:
-
-```bash
-ansible-playbook playbook.yml --tags comfyui-essential
-```
-
-This downloads:
-- Image: FLUX Schnell (23GB) + SDXL Base (7GB)
-- Video: CogVideoX-5B (20GB) + SVD (8GB)
-- Audio: MusicGen Medium (11GB)
-- Support: All 3 CLIP models (11GB)
-- **Total: ~80GB**
-
-### Selective Installation
-
-#### Install ComfyUI Base Only
-
-```bash
-ansible-playbook playbook.yml --tags comfyui
-```
-
-#### Install All Image Models
-
-```bash
-ansible-playbook playbook.yml --tags comfyui-models-image
-```
-
-#### Install All Video Models
-
-```bash
-ansible-playbook playbook.yml --tags comfyui-models-video
-```
-
-#### Install All Audio Models
-
-```bash
-ansible-playbook playbook.yml --tags comfyui-models-audio
-```
-
-#### Install Support Models (CLIP, IP-Adapter)
-
-```bash
-ansible-playbook playbook.yml --tags comfyui-models-support
-```
-
-#### Install Custom Nodes
-
-```bash
-ansible-playbook playbook.yml --tags comfyui-nodes
-```
-
-### Combined Installation
-
-#### Full ComfyUI Setup
-
-```bash
-ansible-playbook playbook.yml --tags comfyui,comfyui-models-all,comfyui-nodes
-```
-
-#### ComfyUI + Image + Video Only
-
-```bash
-ansible-playbook playbook.yml --tags comfyui,comfyui-models-image,comfyui-models-video,comfyui-nodes
-```
-
-## Adding New Models
-
-### Add Image Model
-
-Edit `playbook.yml` and append to `comfyui_image_models`:
-
-```yaml
-- name: "author/model-name"
-  type: "checkpoint"
-  category: "image"
-  size_gb: 15
-  vram_gb: 18
-  format: "fp16"
-  description: "Model description"
-  essential: false  # Set to true for default installs
-```
-
-### Add Custom Node
-
-Append to `comfyui_custom_nodes`:
-
-```yaml
-- name: "NodeName"
-  repo: "https://github.com/author/repo.git"
-  category: "image"  # or video, audio, control, etc.
-  description: "What it does"
-  essential: false
-```
-
-## VRAM Requirements Summary
-
-For your **24GB GPU**, you can run:
-
-### Simultaneously Loadable Combinations
-
-**Option 1: Image Generation Focus**
-- FLUX Schnell FP16 (23GB) → Full model, leaves 1GB for operations
-- FLUX Schnell FP8 (12GB) + SDXL (12GB) → Both loaded
-- SDXL (12GB) + Refiner (12GB) → Both loaded with room
-
-**Option 2: Video Generation**
-- CogVideoX-5B with optimizations (12GB) + SDXL (12GB)
-- SVD-XT (20GB) → Good frame count
-
-**Option 3: Multi-Modal**
-- SDXL (12GB) + MusicGen Medium (8GB)
-- CogVideoX-5B (12GB) + MusicGen Small (4GB)
-
-## Storage Requirements
-
-### Essential Models Only
-- Image: 30GB
-- Video: 28GB
-- Audio: 11GB
-- Support: 11GB
-- **Total: ~80GB**
-
-### All Models (including optional)
-- Image: 54GB
-- Video: 36GB
-- Audio: 36GB
-- Support: 11GB
-- **Total: ~137GB**
-
-## Directory Structure Created
-
-```
-/workspace/ComfyUI/
-├── models/
-│   ├── checkpoints/        # FLUX, SDXL, SD3 models
-│   ├── unet/               # Diffusion U-Nets
-│   ├── vae/                # VAE models
-│   ├── loras/              # LoRA adapters
-│   ├── clip/               # CLIP text encoders
-│   ├── clip_vision/        # CLIP vision models
-│   ├── controlnet/         # ControlNet models
-│   ├── ipadapter/          # IP-Adapter models
-│   ├── embeddings/         # Text embeddings
-│   ├── upscale_models/     # Upscalers
-│   ├── video_models/       # Video generation models
-│   ├── animatediff_models/ # AnimateDiff models
-│   ├── audio_models/       # Audio generation models
-│   └── configs/            # Model configs
-└── custom_nodes/           # Extension nodes
-```
-
-## Extending the Playbook
-
-### Add a New Model Category
-
-1. Add variable list in `vars` section:
-
-```yaml
-comfyui_new_category_models:
-  - name: "model-repo-id"
-    type: "new_type"
-    category: "new_category"
-    size_gb: 10
-    description: "Description"
-    essential: true
-```
-
-2. Create download task section:
-
-```yaml
-- name: Download New Category Models
-  tags: [comfyui-models-newcat, comfyui-models-all]
-  block:
-    # Copy pattern from existing model download tasks
-```
-
-3. Update tags documentation at top of file
-
-## Model Download Notes
-
-- **Async Downloads**: Large models download asynchronously with 1-hour timeouts
-- **Smart Caching**: Models are only downloaded if not already cached
-- **HuggingFace Token**: Set `HF_TOKEN` environment variable for gated models
-- **Error Handling**: Download failures are ignored to allow partial installations
-
-## Performance Tips
-
-### For 24GB GPU:
-
-1. **Use FP8 Quantized FLUX** when you need multiple models loaded
-2. **Use Full FP16 FLUX** when you want maximum quality
-3. **Enable CogVideoX Optimizations** to fit in 12GB (edit model config)
-4. **Use GGUF variants** for FLUX to save VRAM (manually download)
-
-### Recommended First Install:
-
-```bash
-# Essential setup (~80GB, ~1 hour)
-ansible-playbook playbook.yml --tags comfyui-essential
-```
-
-Then add specific models as needed:
-
-```bash
-# Add video models later
-ansible-playbook playbook.yml --tags comfyui-models-video
-```
-
-## Troubleshooting
-
-### Downloads Failing
-
-Check HuggingFace token:
-```bash
-export HF_TOKEN="your_token_here"
-ansible-playbook playbook.yml --tags comfyui-models-image
-```
-
-### Out of Space
-
-Check available storage:
-```bash
-df -h /workspace
-du -sh /workspace/huggingface_cache
-```
-
-Remove optional models:
-```bash
-# Keep only essential models in playbook vars
-```
-
-### Custom Node Dependencies
-
-Some nodes require manual intervention:
-```bash
-cd /workspace/ComfyUI/custom_nodes/NodeName
-pip3 install -r requirements.txt
-```
-
-## Next Steps
-
-After installation:
-
-1. **Start ComfyUI:**
-   ```bash
-   cd /workspace/ComfyUI
-   python3 main.py --listen 0.0.0.0 --port 8188
-   ```
-
-2. **Access via Browser:**
-   ```
-   http://your-runpod-ip:8188
-   ```
-
-3. **Install Additional Nodes via Manager:**
-   - ComfyUI Manager provides GUI for installing more nodes
-   - Browse and install from the manager interface
-
-4. **Download Model Variants:**
-   - Use ComfyUI Manager to download LoRAs
-   - Download ControlNet models as needed
-   - Get additional checkpoints from Civitai
-
-## Support
-
-For issues or questions:
-- Check Ansible logs for download status
-- Verify HuggingFace token is valid
-- Ensure sufficient storage space
-- Check VRAM requirements for your GPU

Dockerfile

@@ -0,0 +1,26 @@
+# RunPod AI Orchestrator Template
+# Minimal Docker image for ComfyUI + vLLM orchestration
+# Models and application code live on network volume at /workspace
+
+FROM runpod/pytorch:2.4.0-py3.11-cuda12.4.1-devel-ubuntu22.04
+
+# Install Supervisor for process management
+RUN pip install --no-cache-dir supervisor
+
+# Install Tailscale for VPN connectivity
+RUN curl -fsSL https://tailscale.com/install.sh | sh
+
+# Install additional system utilities
+RUN apt-get update && apt-get install -y \
+    wget \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy the startup script
+COPY start.sh /start.sh
+RUN chmod +x /start.sh
+
+# Set working directory to /workspace (network volume mount point)
+WORKDIR /workspace
+
+# RunPod calls /start.sh by default
+CMD ["/start.sh"]

RUNPOD_TEMPLATE.md

@@ -0,0 +1,503 @@
+# RunPod Template Setup Guide
+
+This guide explains how to deploy the AI Orchestrator (ComfyUI + vLLM) on RunPod using a custom Docker template and network volume.
+
+## Architecture Overview
+
+The deployment uses a **two-tier strategy**:
+
+1. **Docker Image** (software layer) - Contains system packages, Supervisor, Tailscale
+2. **Network Volume** (data layer) - Contains models, ComfyUI installation, venvs, configuration
+
+This approach allows fast pod deployment (~2-3 minutes) while keeping all large files (models, ~80-200GB) on a persistent network volume.
+
+## Prerequisites
+
+- RunPod account with credits
+- Docker Hub account (for hosting the template image)
+- HuggingFace account with API token (for model downloads)
+- Tailscale account with auth key (optional, for VPN access)
+
+## Step 1: Build and Push Docker Image
+
+### Option A: Automated Build (Recommended)
+
+The repository includes a Gitea workflow that automatically builds and pushes the Docker image to your Gitea container registry when you push to the `main` branch or create a version tag.
+
+1. **Configure Gitea Secret:**
+   - Go to your Gitea repository → Settings → Secrets
+   - Add `REGISTRY_TOKEN` = your Gitea access token with registry permissions
+   - (The workflow automatically uses your Gitea username via `gitea.actor`)
+
+2. **Trigger Build:**
+   ```bash
+   # Push to main branch
+   git push origin main
+
+   # Or create a version tag
+   git tag v1.0.0
+   git push origin v1.0.0
+   ```
+
+3. **Monitor Build:**
+   - Go to Actions tab in Gitea
+   - Wait for build to complete (~5-10 minutes)
+   - Note the Docker image name: `dev.pivoine.art/valknar/runpod-ai-orchestrator:latest`
+
+### Option B: Manual Build
+
+If you prefer to build manually:
+
+```bash
+# From the repository root
+cd /path/to/runpod
+
+# Build the image
+docker build -t dev.pivoine.art/valknar/runpod-ai-orchestrator:latest .
+
+# Login to your Gitea registry
+docker login dev.pivoine.art
+
+# Push to Gitea registry
+docker push dev.pivoine.art/valknar/runpod-ai-orchestrator:latest
+```
+
+## Step 2: Create Network Volume
+
+Network volumes persist your models and data across pod restarts and rebuilds.
+
+1. **Go to RunPod Dashboard → Storage → Network Volumes**
+
+2. **Click "New Network Volume"**
+
+3. **Configure:**
+   - **Name**: `ai-orchestrator-models`
+   - **Size**: `200GB` (adjust based on your needs)
+     - Essential models only: ~80GB
+     - All models: ~137-200GB
+   - **Datacenter**: Choose closest to you (volume tied to datacenter)
+
+4. **Click "Create Volume"**
+
+5. **Note the Volume ID** (e.g., `vol-abc123def456`) for pod deployment
+
+### Storage Requirements
+
+| Configuration | Size | Models Included |
+|--------------|------|-----------------|
+| Essential | ~80GB | FLUX Schnell, 1-2 SDXL checkpoints, MusicGen Medium |
+| Complete | ~137GB | All image/video/audio models from playbook |
+| Full + vLLM | ~200GB | Complete + Qwen 2.5 7B + Llama 3.1 8B |
+
+## Step 3: Create RunPod Template
+
+1. **Go to RunPod Dashboard → Templates**
+
+2. **Click "New Template"**
+
+3. **Configure Template Settings:**
+
+   **Container Configuration:**
+   - **Template Name**: `AI Orchestrator (ComfyUI + vLLM)`
+   - **Template Type**: Docker
+   - **Container Image**: `dev.pivoine.art/valknar/runpod-ai-orchestrator:latest`
+   - **Container Disk**: `50GB` (for system and temp files)
+   - **Docker Command**: Leave empty (uses default `/start.sh`)
+
+   **Volume Configuration:**
+   - **Volume Mount Path**: `/workspace`
+   - **Attach to Network Volume**: Select your volume ID from Step 2
+
+   **Port Configuration:**
+   - **Expose HTTP Ports**: `8188, 9000, 9001`
+     - `8188` - ComfyUI web interface
+     - `9000` - Model orchestrator API
+     - `9001` - Supervisor web UI
+   - **Expose TCP Ports**: `22` (SSH access)
+
+   **Environment Variables:**
+   ```
+   HF_TOKEN=your_huggingface_token_here
+   TAILSCALE_AUTHKEY=tskey-auth-your_tailscale_authkey_here
+   SUPERVISOR_BACKEND_HOST=localhost
+   SUPERVISOR_BACKEND_PORT=9001
+   ```
+
+   **Advanced Settings:**
+   - **Start Jupyter**: No
+   - **Start SSH**: Yes (handled by base image)
+
+4. **Click "Save Template"**
+
+## Step 4: First Deployment (Initial Setup)
+
+The first time you deploy, you need to set up the network volume with models and configuration.
+
+### 4.1 Deploy Pod
+
+1. **Go to RunPod Dashboard → Pods**
+2. **Click "Deploy"** or "GPU Pods"
+3. **Select your custom template**: `AI Orchestrator (ComfyUI + vLLM)`
+4. **Configure GPU:**
+   - **GPU Type**: RTX 4090 (24GB VRAM) or higher
+   - **Network Volume**: Select your volume from Step 2
+   - **On-Demand vs Spot**: Choose based on budget
+5. **Click "Deploy"**
+
+### 4.2 SSH into Pod
+
+```bash
+# Get pod SSH command from RunPod dashboard
+ssh root@<pod-ip> -p <port> -i ~/.ssh/id_ed25519
+
+# Or use RunPod web terminal
+```
+
+### 4.3 Initial Setup on Network Volume
+
+```bash
+# 1. Clone the repository to /workspace/ai
+cd /workspace
+git clone https://github.com/your-username/runpod.git ai
+cd ai
+
+# 2. Create .env file with your credentials
+cp .env.example .env
+nano .env
+
+# Edit and add:
+# HF_TOKEN=your_huggingface_token
+# TAILSCALE_AUTHKEY=tskey-auth-your_key
+# GPU_TAILSCALE_IP=<will be set automatically>
+
+# 3. Download essential models (this takes 30-60 minutes)
+ansible-playbook playbook.yml --tags comfyui-essential
+
+# OR download all models (1-2 hours)
+ansible-playbook playbook.yml --tags comfyui-models-all
+
+# 4. Link models to ComfyUI
+bash scripts/link-comfyui-models.sh
+
+# OR if arty is available
+arty run models/link-comfyui
+
+# 5. Install ComfyUI custom nodes dependencies
+cd /workspace/ComfyUI/custom_nodes/ComfyUI-Manager
+pip install -r requirements.txt
+cd /workspace/ai
+
+# 6. Restart the container to apply all changes
+exit
+# Go to RunPod dashboard → Stop pod → Start pod
+```
+
+### 4.4 Verify Services
+
+After restart, SSH back in and check:
+
+```bash
+# Check supervisor status
+supervisorctl -c /workspace/supervisord.conf status
+
+# Expected output:
+# comfyui                          RUNNING   pid 123, uptime 0:01:00
+# (orchestrator is disabled by default - enable for vLLM)
+
+# Test ComfyUI
+curl -I http://localhost:8188
+
+# Test Supervisor web UI
+curl -I http://localhost:9001
+```
+
+## Step 5: Subsequent Deployments
+
+After initial setup, deploying new pods is quick (2-3 minutes):
+
+1. **Deploy pod** with same template + network volume
+2. **Wait for startup** (~1-2 minutes for services to start)
+3. **Access services:**
+   - ComfyUI: `http://<pod-ip>:8188`
+   - Supervisor: `http://<pod-ip>:9001`
+
+**All models, configuration, and data persist on the network volume!**
+
+## Step 6: Access Services
+
+### Via Direct IP (HTTP)
+
+Get pod IP and ports from RunPod dashboard:
+
+```
+ComfyUI:           http://<pod-ip>:8188
+Supervisor UI:     http://<pod-ip>:9001
+Orchestrator API:  http://<pod-ip>:9000
+SSH:               ssh root@<pod-ip> -p <port>
+```
+
+### Via Tailscale VPN (Recommended)
+
+If you configured `TAILSCALE_AUTHKEY`, the pod automatically joins your Tailscale network:
+
+1. **Get Tailscale IP:**
+   ```bash
+   ssh root@<pod-ip> -p <port>
+   tailscale ip -4
+   # Example output: 100.114.60.40
+   ```
+
+2. **Access via Tailscale:**
+   ```
+   ComfyUI:      http://<tailscale-ip>:8188
+   Supervisor:   http://<tailscale-ip>:9001
+   Orchestrator: http://<tailscale-ip>:9000
+   SSH:          ssh root@<tailscale-ip>
+   ```
+
+3. **Update LiteLLM config** on your VPS with the Tailscale IP
+
+## Service Management
+
+### Start/Stop Services
+
+```bash
+# Start all services
+supervisorctl -c /workspace/supervisord.conf start all
+
+# Stop all services
+supervisorctl -c /workspace/supervisord.conf stop all
+
+# Restart specific service
+supervisorctl -c /workspace/supervisord.conf restart comfyui
+
+# View status
+supervisorctl -c /workspace/supervisord.conf status
+```
+
+### Enable vLLM Models (Text Generation)
+
+By default, only ComfyUI runs (to save VRAM). To enable vLLM:
+
+1. **Stop ComfyUI** (frees up VRAM):
+   ```bash
+   supervisorctl -c /workspace/supervisord.conf stop comfyui
+   ```
+
+2. **Start orchestrator** (manages vLLM models):
+   ```bash
+   supervisorctl -c /workspace/supervisord.conf start orchestrator
+   ```
+
+3. **Test text generation:**
+   ```bash
+   curl -X POST http://localhost:9000/v1/chat/completions \
+     -H 'Content-Type: application/json' \
+     -d '{"model":"qwen-2.5-7b","messages":[{"role":"user","content":"Hello"}]}'
+   ```
+
+### Switch Back to ComfyUI
+
+```bash
+# Stop orchestrator (stops all vLLM models)
+supervisorctl -c /workspace/supervisord.conf stop orchestrator
+
+# Start ComfyUI
+supervisorctl -c /workspace/supervisord.conf start comfyui
+```
+
+## Updating the Template
+
+When you make changes to code or configuration:
+
+### Update Docker Image
+
+```bash
+# 1. Make changes to Dockerfile or start.sh
+# 2. Push to repository
+git add .
+git commit -m "Update template configuration"
+git push origin main
+
+# 3. Gitea workflow auto-builds new image
+
+# 4. Terminate old pod and deploy new one with updated image
+```
+
+### Update Network Volume Data
+
+```bash
+# SSH into running pod
+ssh root@<pod-ip> -p <port>
+
+# Update repository
+cd /workspace/ai
+git pull
+
+# Re-run Ansible if needed
+ansible-playbook playbook.yml --tags <specific-tag>
+
+# Restart services
+supervisorctl -c /workspace/supervisord.conf restart all
+```
+
+## Troubleshooting
+
+### Pod fails to start
+
+**Check logs:**
+```bash
+# Via SSH
+cat /workspace/logs/supervisord.log
+cat /workspace/logs/comfyui.err.log
+
+# Via RunPod web terminal
+tail -f /workspace/logs/*.log
+```
+
+**Common issues:**
+- Missing `.env` file → Create `/workspace/ai/.env` with required vars
+- Supervisor config not found → Ensure `/workspace/ai/supervisord.conf` exists
+- Port conflicts → Check if services are already running
+
+### Tailscale not connecting
+
+**Check Tailscale status:**
+```bash
+tailscale status
+tailscale ip -4
+```
+
+**Common issues:**
+- Missing or invalid `TAILSCALE_AUTHKEY` in `.env`
+- Auth key expired → Generate new key in Tailscale admin
+- Firewall blocking → RunPod should allow Tailscale by default
+
+### Services not starting
+
+**Check Supervisor:**
+```bash
+supervisorctl -c /workspace/supervisord.conf status
+supervisorctl -c /workspace/supervisord.conf tail -f comfyui
+```
+
+**Common issues:**
+- venv broken → Re-run `scripts/bootstrap-venvs.sh`
+- Models not downloaded → Run Ansible playbook again
+- Python version mismatch → Rebuild venvs
+
+### Out of VRAM
+
+**Check GPU memory:**
+```bash
+nvidia-smi
+```
+
+**RTX 4090 (24GB) capacity:**
+- ComfyUI (FLUX Schnell): ~23GB (can't run with vLLM)
+- vLLM (Qwen 2.5 7B): ~14GB
+- vLLM (Llama 3.1 8B): ~17GB
+
+**Solution:** Only run one service at a time (see Service Management section)
+
+### Network volume full
+
+**Check disk usage:**
+```bash
+df -h /workspace
+du -sh /workspace/*
+```
+
+**Clean up:**
+```bash
+# Remove old HuggingFace cache
+rm -rf /workspace/huggingface_cache
+
+# Re-download essential models only
+cd /workspace/ai
+ansible-playbook playbook.yml --tags comfyui-essential
+```
+
+## Cost Optimization
+
+### Spot vs On-Demand
+
+- **Spot instances**: ~70% cheaper, can be interrupted
+- **On-Demand**: More expensive, guaranteed availability
+
+**Recommendation:** Use spot for development, on-demand for production
+
+### Network Volume Pricing
+
+- First 1TB: $0.07/GB/month
+- Beyond 1TB: $0.05/GB/month
+
+**200GB volume cost:** ~$14/month
+
+### Pod Auto-Stop
+
+Configure auto-stop in RunPod pod settings to save costs when idle:
+- Stop after 15 minutes idle
+- Stop after 1 hour idle
+- Manual stop only
+
+## Advanced Configuration
+
+### Custom Environment Variables
+
+Add to template or pod environment variables:
+
+```bash
+# Model cache locations
+HF_HOME=/workspace/huggingface_cache
+TRANSFORMERS_CACHE=/workspace/huggingface_cache
+
+# ComfyUI settings
+COMFYUI_PORT=8188
+COMFYUI_LISTEN=0.0.0.0
+
+# Orchestrator settings
+ORCHESTRATOR_PORT=9000
+
+# GPU settings
+CUDA_VISIBLE_DEVICES=0
+```
+
+### Multiple Network Volumes
+
+You can attach multiple network volumes for organization:
+
+1. **Models volume** - `/workspace/models` (read-only, shared)
+2. **Data volume** - `/workspace/data` (read-write, per-project)
+
+### Custom Startup Script
+
+Override `/start.sh` behavior by creating `/workspace/custom-start.sh`:
+
+```bash
+#!/bin/bash
+# Custom startup commands
+
+# Source default startup
+source /start.sh
+
+# Add your custom commands here
+echo "Running custom initialization..."
+```
+
+## References
+
+- [RunPod Documentation](https://docs.runpod.io/)
+- [RunPod Templates Overview](https://docs.runpod.io/pods/templates/overview)
+- [Network Volumes Guide](https://docs.runpod.io/storage/network-volumes)
+- [ComfyUI Documentation](https://github.com/comfyanonymous/ComfyUI)
+- [Supervisor Documentation](http://supervisord.org/)
+- [Tailscale Documentation](https://tailscale.com/kb/)
+
+## Support
+
+For issues or questions:
+- Check troubleshooting section above
+- Review `/workspace/logs/` files
+- Check RunPod community forums
+- Open issue in project repository

arty.yml

@@ -63,11 +63,41 @@ references:
     description: "MusicGen and Stable Audio integration"
     essential: false
 
+  - url: https://github.com/billwuhao/ComfyUI_DiffRhythm.git
+    into: $COMFYUI_ROOT/custom_nodes/ComfyUI_DiffRhythm
+    description: "DiffRhythm - Full-length song generation (up to 4m45s) with text/audio conditioning"
+    essential: false
+
+  - url: https://github.com/billwuhao/ComfyUI_ACE-Step.git
+    into: $COMFYUI_ROOT/custom_nodes/ComfyUI_ACE-Step
+    description: "ACE Step - State-of-the-art music generation with 19-language support, voice cloning, and superior coherence"
+    essential: false
+
   - url: https://github.com/ssitu/ComfyUI_UltimateSDUpscale.git
     into: $COMFYUI_ROOT/custom_nodes/ComfyUI_UltimateSDUpscale
     description: "Ultimate SD Upscale for high-quality image upscaling"
     essential: false
 
+  - url: https://github.com/kijai/ComfyUI-KJNodes.git
+    into: $COMFYUI_ROOT/custom_nodes/ComfyUI-KJNodes
+    description: "Kijai optimizations for HunyuanVideo and Wan2.2 (FP8 scaling, video helpers, model loading)"
+    essential: true
+
+  - url: https://github.com/Fannovel16/comfyui_controlnet_aux.git
+    into: $COMFYUI_ROOT/custom_nodes/comfyui_controlnet_aux
+    description: "ControlNet preprocessors (Canny, Depth, OpenPose, MLSD) for Wan2.2 Fun Control"
+    essential: true
+
+  - url: https://github.com/city96/ComfyUI-GGUF.git
+    into: $COMFYUI_ROOT/custom_nodes/ComfyUI-GGUF
+    description: "GGUF quantization support for memory-efficient model loading"
+    essential: false
+
+  - url: https://github.com/11cafe/comfyui-workspace-manager.git
+    into: $COMFYUI_ROOT/custom_nodes/comfyui-workspace-manager
+    description: "Workspace manager for ComfyUI - workflow/model organization (obsolete but requested)"
+    essential: false
+
 # Environment profiles for selective repository management
 envs:
   # RunPod environment variables
@@ -78,37 +108,6 @@ envs:
     LOGS_DIR: /workspace/logs
     BIN_DIR: /workspace/bin
 
-  # Production: Only essential components
-  prod:
-    - $AI_ROOT
-    - $COMFYUI_ROOT
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Manager
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-VideoHelperSuite
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-AnimateDiff-Evolved
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI_IPAdapter_plus
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Impact-Pack
-
-  # Development: All repositories including optional nodes
-  dev:
-    - $AI_ROOT
-    - $COMFYUI_ROOT
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Manager
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-VideoHelperSuite
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-AnimateDiff-Evolved
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI_IPAdapter_plus
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Impact-Pack
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-CogVideoXWrapper
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Inspire-Pack
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Advanced-ControlNet
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-3D-Pack
-    - $COMFYUI_ROOT/custom_nodes/comfyui-sound-lab
-
-  # Minimal: Only orchestrator and ComfyUI base
-  minimal:
-    - $AI_ROOT
-    - $COMFYUI_ROOT
-    - $COMFYUI_ROOT/custom_nodes/ComfyUI-Manager
-
 # Deployment scripts for RunPod instances
 scripts:
   #
@@ -165,11 +164,23 @@ scripts:
       htop \
       tmux \
       net-tools \
-      davfs2
+      davfs2 \
+      ffmpeg \
+      libavcodec-dev \
+      libavformat-dev \
+      libavutil-dev \
+      libswscale-dev
 
     echo ""
     echo "✓ System packages installed successfully"
 
+    # Verify FFmpeg installation
+    if ffmpeg -version > /dev/null 2>&1; then
+      echo "✓ FFmpeg installed: $(ffmpeg -version | head -1 | cut -d ' ' -f3)"
+    else
+      echo "❌ WARNING: FFmpeg not found"
+    fi
+
   setup/python-env: |
     echo "========================================="
     echo "  Setting Up Python Environment"
@@ -279,43 +290,67 @@ scripts:
     echo "========================================="
     echo ""
 
+    # Install system dependencies
+    echo "Installing system dependencies..."
+    sudo apt-get update -qq
+    sudo apt-get install -y -qq espeak-ng
+    echo "✓ System dependencies installed (espeak-ng)"
+    echo ""
+
     cd $COMFYUI_ROOT/custom_nodes
 
     # ComfyUI Manager
-    echo "[1/5] Installing ComfyUI-Manager..."
+    echo "[1/6] Installing ComfyUI-Manager..."
     if [ ! -d "ComfyUI-Manager" ]; then
       git clone https://github.com/ltdrdata/ComfyUI-Manager.git
     fi
     [ -f "ComfyUI-Manager/requirements.txt" ] && sudo pip3 install -r ComfyUI-Manager/requirements.txt
 
     # VideoHelperSuite
-    echo "[2/5] Installing ComfyUI-VideoHelperSuite..."
+    echo "[2/6] Installing ComfyUI-VideoHelperSuite..."
     if [ ! -d "ComfyUI-VideoHelperSuite" ]; then
       git clone https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite.git
     fi
     [ -f "ComfyUI-VideoHelperSuite/requirements.txt" ] && sudo pip3 install -r ComfyUI-VideoHelperSuite/requirements.txt
 
     # AnimateDiff-Evolved
-    echo "[3/5] Installing ComfyUI-AnimateDiff-Evolved..."
+    echo "[3/6] Installing ComfyUI-AnimateDiff-Evolved..."
     if [ ! -d "ComfyUI-AnimateDiff-Evolved" ]; then
       git clone https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved.git
     fi
     [ -f "ComfyUI-AnimateDiff-Evolved/requirements.txt" ] && sudo pip3 install -r ComfyUI-AnimateDiff-Evolved/requirements.txt
 
     # IPAdapter Plus
-    echo "[4/5] Installing ComfyUI_IPAdapter_plus..."
+    echo "[4/6] Installing ComfyUI_IPAdapter_plus..."
     if [ ! -d "ComfyUI_IPAdapter_plus" ]; then
       git clone https://github.com/cubiq/ComfyUI_IPAdapter_plus.git
     fi
     [ -f "ComfyUI_IPAdapter_plus/requirements.txt" ] && sudo pip3 install -r ComfyUI_IPAdapter_plus/requirements.txt
 
     # Impact-Pack
-    echo "[5/5] Installing ComfyUI-Impact-Pack..."
+    echo "[5/6] Installing ComfyUI-Impact-Pack..."
     if [ ! -d "ComfyUI-Impact-Pack" ]; then
       git clone https://github.com/ltdrdata/ComfyUI-Impact-Pack.git
     fi
     [ -f "ComfyUI-Impact-Pack/requirements.txt" ] && sudo pip3 install -r ComfyUI-Impact-Pack/requirements.txt
 
+    # DiffRhythm
+    echo "[6/6] Installing ComfyUI_DiffRhythm..."
+    if [ ! -d "ComfyUI_DiffRhythm" ]; then
+      git clone https://github.com/billwuhao/ComfyUI_DiffRhythm.git
+    fi
+    if [ -f "ComfyUI_DiffRhythm/requirements.txt" ]; then
+      cd $COMFYUI_ROOT
+      source venv/bin/activate
+      pip install -r custom_nodes/ComfyUI_DiffRhythm/requirements.txt
+      deactivate
+      cd custom_nodes
+    fi
+
+    # Create DiffRhythm model directories
+    echo "Creating DiffRhythm model directories..."
+    mkdir -p $COMFYUI_ROOT/models/TTS/DiffRhythm/{MuQ-large-msd-iter,MuQ-MuLan-large,xlm-roberta-base,eval-model}
+
     # Fix numpy version for vLLM compatibility
     echo "Fixing numpy version..."
     sudo pip3 install 'numpy<2.0.0' --force-reinstall
@@ -327,6 +362,144 @@ scripts:
     echo "  - AnimateDiff-Evolved: Video generation"
     echo "  - IPAdapter_plus: Style transfer"
     echo "  - Impact-Pack: Face enhancement"
+    echo "  - DiffRhythm: Full-length song generation"
+
+  models/diffrhythm-eval: |
+    echo "========================================="
+    echo "  Downloading DiffRhythm Eval Model"
+    echo "========================================="
+    echo ""
+
+    # Create eval-model directory
+    mkdir -p $COMFYUI_ROOT/models/TTS/DiffRhythm/eval-model
+    cd $COMFYUI_ROOT/models/TTS/DiffRhythm/eval-model
+
+    # Download eval.yaml (129 bytes)
+    echo "Downloading eval.yaml..."
+    curl -L -o eval.yaml "https://huggingface.co/spaces/ASLP-lab/DiffRhythm/resolve/main/pretrained/eval.yaml"
+
+    # Download eval.safetensors (101 MB)
+    echo "Downloading eval.safetensors (101 MB)..."
+    curl -L -o eval.safetensors "https://huggingface.co/spaces/ASLP-lab/DiffRhythm/resolve/main/pretrained/eval.safetensors"
+
+    # Verify files
+    if [ -f "eval.yaml" ] && [ -f "eval.safetensors" ]; then
+      echo ""
+      echo "✓ DiffRhythm eval-model files downloaded successfully"
+      echo "  - eval.yaml: $(du -h eval.yaml | cut -f1)"
+      echo "  - eval.safetensors: $(du -h eval.safetensors | cut -f1)"
+    else
+      echo "❌ ERROR: Failed to download eval-model files"
+      exit 1
+    fi
+
+  setup/comfyui-acestep: |
+    echo "========================================="
+    echo "  Installing ACE Step Custom Node"
+    echo "========================================="
+    echo ""
+
+    cd $COMFYUI_ROOT/custom_nodes
+
+    # Clone repository if not exists
+    if [ ! -d "ComfyUI_ACE-Step" ]; then
+      echo "Cloning ComfyUI_ACE-Step repository..."
+      git clone https://github.com/billwuhao/ComfyUI_ACE-Step.git
+    else
+      echo "ComfyUI_ACE-Step already exists, skipping clone"
+    fi
+
+    # Install dependencies in ComfyUI venv
+    echo ""
+    echo "Installing ACE Step dependencies..."
+    cd $COMFYUI_ROOT
+    source venv/bin/activate
+    pip install -r custom_nodes/ComfyUI_ACE-Step/requirements.txt
+    deactivate
+
+    echo ""
+    echo "✓ ACE Step custom node installed successfully"
+    echo "  Note: Download models separately using:"
+    echo "  bash /workspace/bin/artifact_huggingface_download.sh download -c models_huggingface.yaml --category audio_models"
+
+  setup/pivoine-nodes: |
+    echo "========================================="
+    echo "  Linking Pivoine Custom Nodes"
+    echo "========================================="
+    echo ""
+
+    NODES_SRC="/workspace/ai/comfyui/nodes"
+    NODES_DEST="/workspace/ComfyUI/custom_nodes/ComfyUI_Pivoine"
+
+    # Remove existing symlink if present
+    if [ -L "$NODES_DEST" ] || [ -d "$NODES_DEST" ]; then
+      echo "Removing existing: $NODES_DEST"
+      rm -rf "$NODES_DEST"
+    fi
+
+    # Create symlink
+    ln -s "$NODES_SRC" "$NODES_DEST"
+
+    echo ""
+    echo "✓ Pivoine custom nodes linked"
+    echo "  Source: $NODES_SRC"
+    echo "  Linked: $NODES_DEST"
+    echo ""
+    echo "Available Pivoine nodes:"
+    echo "  🌸 PivoineDiffRhythmRun - DiffRhythm with chunked disabled"
+    echo ""
+    echo "Category: 🌸Pivoine/Audio"
+
+  fix/diffrhythm-patch: |
+    echo "========================================="
+    echo "  Apply DiffRhythm LlamaConfig Patch"
+    echo "========================================="
+    echo ""
+    echo "Issue: Tensor dimension mismatch (32 vs 64) in rotary embeddings"
+    echo "Solution: Patch DiffRhythm __init__.py to fix LlamaConfig"
+    echo ""
+    echo "References:"
+    echo "  - https://github.com/billwuhao/ComfyUI_DiffRhythm/issues/44"
+    echo "  - https://github.com/billwuhao/ComfyUI_DiffRhythm/issues/48"
+    echo ""
+
+    DIFF_RHYTHM_DIR="/workspace/ComfyUI/custom_nodes/ComfyUI_DiffRhythm"
+    PATCH_FILE="/workspace/ai/comfyui/patches/diffrhythm-llamaconfig-fix.patch"
+
+    if [ ! -d "$DIFF_RHYTHM_DIR" ]; then
+      echo "✗ Error: DiffRhythm not found at $DIFF_RHYTHM_DIR"
+      exit 1
+    fi
+
+    if [ ! -f "$PATCH_FILE" ]; then
+      echo "✗ Error: Patch file not found at $PATCH_FILE"
+      exit 1
+    fi
+
+    cd "$DIFF_RHYTHM_DIR"
+
+    echo "Checking if patch already applied..."
+    if grep -q "PatchedLlamaConfig" __init__.py; then
+      echo "✓ Patch already applied!"
+      exit 0
+    fi
+
+    echo "Applying patch..."
+    patch -p1 < "$PATCH_FILE"
+
+    if [ $? -eq 0 ]; then
+      echo ""
+      echo "✓ Patch applied successfully!"
+      echo ""
+      echo "Next steps:"
+      echo "  1. Restart ComfyUI: arty services/comfyui/restart"
+      echo "  2. Test DiffRhythm workflows"
+    else
+      echo ""
+      echo "✗ Failed to apply patch"
+      echo "You may need to manually apply the patch or check for conflicts"
+      exit 1
+    fi
 
   setup/comfyui-extensions-deps: |
     echo "========================================="
@@ -436,58 +609,6 @@ scripts:
     echo "To manage: supervisorctl status"
     echo "Web UI: http://localhost:9001 (admin/runpod2024)"
 
-  setup/webdav: |
-    echo "========================================="
-    echo "  Setting Up WebDAV Mount (HiDrive)"
-    echo "========================================="
-    echo ""
-
-    # Install davfs2 if not present
-    if ! command -v mount.davfs >/dev/null 2>&1; then
-      echo "Installing davfs2..."
-      DEBIAN_FRONTEND=noninteractive apt update && DEBIAN_FRONTEND=noninteractive apt install -y davfs2
-    fi
-
-    # Create mount point
-    echo "Creating mount point..."
-    mkdir -p /mnt/hidrive
-
-    # Create davfs2 secrets file
-    echo "Configuring WebDAV credentials..."
-    mkdir -p /etc/davfs2
-    echo "https://webdav.hidrive.ionos.com/ valknar MwRTW4hR.eRbipQ" | tee /etc/davfs2/secrets > /dev/null
-    chmod 600 /etc/davfs2/secrets
-
-    # Configure davfs2
-    sed -i 's/# use_locks       1/use_locks       0/' /etc/davfs2/davfs2.conf 2>/dev/null || true
-
-    # Mount WebDAV
-    echo "Mounting HiDrive WebDAV..."
-    if mount -t davfs https://webdav.hidrive.ionos.com/ /mnt/hidrive; then
-      echo "✓ HiDrive mounted successfully"
-    else
-      echo "⚠ Warning: Mount failed, you may need to mount manually"
-      echo "  Try: mount -t davfs https://webdav.hidrive.ionos.com/ /mnt/hidrive"
-    fi
-
-    # Create ComfyUI output directory
-    echo "Creating ComfyUI output directory..."
-    mkdir -p /mnt/hidrive/users/valknar/Pictures/AI/ComfyUI
-
-    # Create symlink in ComfyUI
-    echo "Creating symlink in ComfyUI..."
-    ln -sf /mnt/hidrive/users/valknar/Pictures/AI/ComfyUI $COMFYUI_ROOT/output_hidrive
-
-    echo ""
-    echo "✓ WebDAV setup complete"
-    echo ""
-    echo "Mount point: /mnt/hidrive"
-    echo "ComfyUI output: /mnt/hidrive/users/valknar/Pictures/AI/ComfyUI"
-    echo "ComfyUI symlink: $COMFYUI_ROOT/output_hidrive"
-    echo ""
-    echo "To unmount: umount /mnt/hidrive"
-    echo "To remount: mount -t davfs https://webdav.hidrive.ionos.com/ /mnt/hidrive"
-
   #
   # Utility Scripts
   #
@@ -575,53 +696,6 @@ scripts:
     echo "  3. Name: multi-modal-ai-v2.0"
     echo "  4. Save and test deployment"
 
-  #
-  # Orchestration Scripts
-  #
-  install/minimal: |
-    echo "========================================="
-    echo "  Minimal Installation"
-    echo "========================================="
-    echo ""
-    echo "Installing: System + Python + ComfyUI + Supervisor"
-    echo ""
-
-    arty run setup/system-packages && \
-    arty run setup/python-env && \
-    arty run setup/comfyui-base && \
-    arty run setup/supervisor
-
-    echo ""
-    echo "✓ Minimal installation complete"
-    echo ""
-    echo "Next steps:"
-    echo "  1. Download models: Use Ansible playbook"
-    echo "  2. Link models: arty run models/link-comfyui"
-    echo "  3. Start services: arty run services/start"
-
-  install/essential: |
-    echo "========================================="
-    echo "  Essential Installation"
-    echo "========================================="
-    echo ""
-    echo "Installing: System + Python + ComfyUI + Nodes + Supervisor"
-    echo ""
-
-    arty run setup/system-packages && \
-    arty run setup/python-env && \
-    arty run setup/comfyui-base && \
-    arty run setup/comfyui-nodes && \
-    arty run setup/supervisor
-
-    echo ""
-    echo "✓ Essential installation complete"
-    echo ""
-    echo "Next steps:"
-    echo "  1. Download models: ansible-playbook playbook.yml --tags comfyui-essential"
-    echo "  2. Link models: arty run models/link-comfyui"
-    echo "  3. Link workflows: arty run workflows/link-comfyui"
-    echo "  4. Start services: arty run services/start"
-
   install/full: |
     echo "========================================="
     echo "  Full Installation"
@@ -647,39 +721,6 @@ scripts:
     echo "  4. Configure Tailscale (see instructions above)"
     echo "  5. Start services: arty run services/start"
 
-  #
-  # Legacy Setup (deprecated - use install/* instead)
-  #
-  setup/full-legacy: |
-    cd $AI_ROOT
-    cp .env.example .env
-    echo "⚠ DEPRECATED: Use 'arty run install/full' instead"
-    echo "Edit .env and set HF_TOKEN, then run: ansible-playbook playbook.yml"
-
-  setup/essential-legacy: |
-    cd $AI_ROOT
-    cp .env.example .env
-    echo "⚠ DEPRECATED: Use 'arty run install/essential' instead"
-    echo "Edit .env and set HF_TOKEN, then run: ansible-playbook playbook.yml --tags comfyui-essential"
-
-  # Model linking (run after models are downloaded)
-  models/link-comfyui: |
-    cd $COMFYUI_ROOT/models/diffusers
-    ln -sf $HF_CACHE/models--black-forest-labs--FLUX.1-schnell FLUX.1-schnell
-    ln -sf $HF_CACHE/models--black-forest-labs--FLUX.1-dev FLUX.1-dev
-    ln -sf $HF_CACHE/models--stabilityai--stable-diffusion-xl-base-1.0 stable-diffusion-xl-base-1.0
-    ln -sf $HF_CACHE/models--stabilityai--stable-diffusion-xl-refiner-1.0 stable-diffusion-xl-refiner-1.0
-    ln -sf $HF_CACHE/models--stabilityai--stable-diffusion-3.5-large stable-diffusion-3.5-large
-    cd $COMFYUI_ROOT/models/clip_vision
-    ln -sf $HF_CACHE/models--openai--clip-vit-large-patch14 clip-vit-large-patch14
-    ln -sf $HF_CACHE/models--laion--CLIP-ViT-bigG-14-laion2B-39B-b160k CLIP-ViT-bigG-14
-    ln -sf $HF_CACHE/models--google--siglip-so400m-patch14-384 siglip-so400m-patch14-384
-    cd $COMFYUI_ROOT/models/diffusion_models
-    ln -sf $HF_CACHE/models--THUDM--CogVideoX-5b CogVideoX-5b
-    ln -sf $HF_CACHE/models--stabilityai--stable-video-diffusion-img2vid stable-video-diffusion-img2vid
-    ln -sf $HF_CACHE/models--stabilityai--stable-video-diffusion-img2vid-xt stable-video-diffusion-img2vid-xt
-    echo "Models linked to ComfyUI"
-
   # Workflow linking (link production workflows with category prefixes)
   workflows/link-comfyui: |
     # Create ComfyUI user workflows directory
@@ -743,6 +784,14 @@ scripts:
       fi
     done
 
+    # NSFW workflows (prefix: nsfw_)
+    for file in "$SOURCE_DIR/nsfw"/*.json; do
+      if [ -f "$file" ]; then
+        basename=$(basename "$file")
+        ln -sf "$file" "nsfw_${basename}"
+      fi
+    done
+
     # Count workflows
     WORKFLOW_COUNT=$(ls -1 *.json 2>/dev/null | wc -l)
 
@@ -751,51 +800,80 @@ scripts:
     echo "  Location: $COMFYUI_ROOT/user/default/workflows/"
     echo ""
     echo "Categories (by prefix):"
-    echo "  - t2i_   : 4 text-to-image workflows (FLUX, SDXL, SD3.5)"
+    echo "  - t2i_    : 5 text-to-image workflows (FLUX, SDXL, SD3.5, LoRA Fusion)"
-    echo "  - i2i_   : 3 image-to-image workflows (IP-Adapter)"
+    echo "  - i2i_    : 3 image-to-image workflows (IP-Adapter)"
-    echo "  - i2v_   : 3 image-to-video workflows (CogVideoX, SVD)"
+    echo "  - i2v_    : 3 image-to-video workflows (CogVideoX, SVD)"
-    echo "  - t2m_   : 4 text-to-music workflows (MusicGen)"
+    echo "  - t2m_    : 4 text-to-music workflows (MusicGen)"
-    echo "  - upscale_: 3 upscaling workflows"
+    echo "  - upscale_: 3 upscaling workflows (Ultimate SD, Simple, Face)"
-    echo "  - adv_   : 3 advanced workflows (ControlNet, AnimateDiff, Batch)"
+    echo "  - adv_    : 3 advanced workflows (ControlNet, AnimateDiff, Batch)"
+    echo "  - nsfw_   : 4 NSFW workflows (LUSTIFY, Pony, RealVisXL, Ultimate Upscale)"
     echo ""
+    echo "Total: 25 production workflows"
     echo "Workflows will appear in ComfyUI's workflow browser, sorted by category prefix"
 
   #
   # Service Management (Supervisor-based)
   #
   # All services
-  services/start: supervisorctl -c /workspace/supervisord.conf start ai-services:*
+  services/start: supervisorctl -c /workspace/supervisord.conf start all
-  services/stop: supervisorctl -c /workspace/supervisord.conf stop ai-services:*
+  services/stop: supervisorctl -c /workspace/supervisord.conf stop all
-  services/restart: supervisorctl -c /workspace/supervisord.conf restart ai-services:*
+  services/restart: supervisorctl -c /workspace/supervisord.conf restart all
   services/status: supervisorctl -c /workspace/supervisord.conf status
 
-  # ComfyUI service
+  # ComfyUI services group
-  services/comfyui/start: supervisorctl -c /workspace/supervisord.conf start ai-services:comfyui
+  services/comfyui-group/start: supervisorctl -c /workspace/supervisord.conf start comfyui-services:*
-  services/comfyui/stop: supervisorctl -c /workspace/supervisord.conf stop ai-services:comfyui
+  services/comfyui-group/stop: supervisorctl -c /workspace/supervisord.conf stop comfyui-services:*
-  services/comfyui/restart: supervisorctl -c /workspace/supervisord.conf restart ai-services:comfyui
+  services/comfyui-group/restart: supervisorctl -c /workspace/supervisord.conf restart comfyui-services:*
-  services/comfyui/status: supervisorctl -c /workspace/supervisord.conf status ai-services:comfyui
+  services/comfyui-group/status: supervisorctl -c /workspace/supervisord.conf status comfyui-services:*
-  services/comfyui/logs: supervisorctl -c /workspace/supervisord.conf tail -f ai-services:comfyui
 
-  # Orchestrator service
+  # vLLM services group
-  services/orchestrator/start: supervisorctl -c /workspace/supervisord.conf start ai-services:orchestrator
+  services/vllm-group/start: supervisorctl -c /workspace/supervisord.conf start vllm-services:*
-  services/orchestrator/stop: supervisorctl -c /workspace/supervisord.conf stop ai-services:orchestrator
+  services/vllm-group/stop: supervisorctl -c /workspace/supervisord.conf stop vllm-services:*
-  services/orchestrator/restart: supervisorctl -c /workspace/supervisord.conf restart ai-services:orchestrator
+  services/vllm-group/restart: supervisorctl -c /workspace/supervisord.conf restart vllm-services:*
-  services/orchestrator/status: supervisorctl -c /workspace/supervisord.conf status ai-services:orchestrator
+  services/vllm-group/status: supervisorctl -c /workspace/supervisord.conf status vllm-services:*
-  services/orchestrator/logs: supervisorctl -c /workspace/supervisord.conf tail -f ai-services:orchestrator
+
+  # ComfyUI service
+  services/comfyui/start: supervisorctl -c /workspace/supervisord.conf start comfyui-services:comfyui
+  services/comfyui/stop: supervisorctl -c /workspace/supervisord.conf stop comfyui-services:comfyui
+  services/comfyui/restart: supervisorctl -c /workspace/supervisord.conf restart comfyui-services:comfyui
+  services/comfyui/status: supervisorctl -c /workspace/supervisord.conf status comfyui-services:comfyui
+  services/comfyui/logs: supervisorctl -c /workspace/supervisord.conf tail -f comfyui-services:comfyui
 
   # WebDAV Sync service
-  services/webdav-sync/start: supervisorctl -c /workspace/supervisord.conf start ai-services:webdav-sync
+  services/webdav-sync/start: supervisorctl -c /workspace/supervisord.conf start comfyui-services:webdav-sync
-  services/webdav-sync/stop: supervisorctl -c /workspace/supervisord.conf stop ai-services:webdav-sync
+  services/webdav-sync/stop: supervisorctl -c /workspace/supervisord.conf stop comfyui-services:webdav-sync
-  services/webdav-sync/restart: supervisorctl -c /workspace/supervisord.conf restart ai-services:webdav-sync
+  services/webdav-sync/restart: supervisorctl -c /workspace/supervisord.conf restart comfyui-services:webdav-sync
-  services/webdav-sync/status: supervisorctl -c /workspace/supervisord.conf status ai-services:webdav-sync
+  services/webdav-sync/status: supervisorctl -c /workspace/supervisord.conf status comfyui-services:webdav-sync
-  services/webdav-sync/logs: supervisorctl -c /workspace/supervisord.conf tail -f ai-services:webdav-sync
+  services/webdav-sync/logs: supervisorctl -c /workspace/supervisord.conf tail -f comfyui-services:webdav-sync
+
+  # vLLM Qwen service
+  services/vllm-qwen/start: supervisorctl -c /workspace/supervisord.conf start vllm-services:vllm-qwen
+  services/vllm-qwen/stop: supervisorctl -c /workspace/supervisord.conf stop vllm-services:vllm-qwen
+  services/vllm-qwen/restart: supervisorctl -c /workspace/supervisord.conf restart vllm-services:vllm-qwen
+  services/vllm-qwen/status: supervisorctl -c /workspace/supervisord.conf status vllm-services:vllm-qwen
+  services/vllm-qwen/logs: supervisorctl -c /workspace/supervisord.conf tail -f vllm-services:vllm-qwen
+
+  # vLLM Llama service
+  services/vllm-llama/start: supervisorctl -c /workspace/supervisord.conf start vllm-services:vllm-llama
+  services/vllm-llama/stop: supervisorctl -c /workspace/supervisord.conf stop vllm-services:vllm-llama
+  services/vllm-llama/restart: supervisorctl -c /workspace/supervisord.conf restart vllm-services:vllm-llama
+  services/vllm-llama/status: supervisorctl -c /workspace/supervisord.conf status vllm-services:vllm-llama
+  services/vllm-llama/logs: supervisorctl -c /workspace/supervisord.conf tail -f vllm-services:vllm-llama
+
+  # vLLM Embedding service
+  services/vllm-embedding/start: supervisorctl -c /workspace/supervisord.conf start vllm-services:vllm-embedding
+  services/vllm-embedding/stop: supervisorctl -c /workspace/supervisord.conf stop vllm-services:vllm-embedding
+  services/vllm-embedding/restart: supervisorctl -c /workspace/supervisord.conf restart vllm-services:vllm-embedding
+  services/vllm-embedding/status: supervisorctl -c /workspace/supervisord.conf status vllm-services:vllm-embedding
+  services/vllm-embedding/logs: supervisorctl -c /workspace/supervisord.conf tail -f vllm-services:vllm-embedding
 
   #
   # Health Checks
   #
-  health/orchestrator: curl http://localhost:9000/health
   health/comfyui: curl http://localhost:8188
-  health/vllm: curl http://localhost:8000/health
+  health/vllm-qwen: curl http://localhost:8000/health
+  health/vllm-llama: curl http://localhost:8001/health
+  health/vllm-embedding: curl http://localhost:8002/health
 
   #
   # System Checks
@@ -913,15 +991,16 @@ notes: |
 
   Link workflows: arty run workflows/link-comfyui
 
-  20 Production Workflows:
+  25 Production Workflows:
-  - Text-to-Image: FLUX Schnell, FLUX Dev, SDXL+Refiner, SD3.5
+  - Text-to-Image: FLUX Schnell, FLUX Dev, SDXL+Refiner, SD3.5, LoRA Fusion
   - Image-to-Image: IP-Adapter (style, face, composition)
   - Image-to-Video: CogVideoX, SVD, SVD-XT
   - Text-to-Music: MusicGen (small/medium/large/melody)
   - Upscaling: Ultimate SD, Simple, Face-focused
   - Advanced: ControlNet, AnimateDiff, Batch processing
+  - NSFW: LUSTIFY Realistic, Pony Diffusion, RealVisXL Lightning, Ultimate Upscale
 
-  Documentation: README.md, WORKFLOW_STANDARDS.md
+  Documentation: README.md, WORKFLOW_STANDARDS.md, nsfw/README.md
 
   ========================================
   ENVIRONMENT PROFILES

comfyui/fix_workflows.py

@@ -1,309 +0,0 @@
-#!/usr/bin/env python3
-"""
-ComfyUI Workflow Schema Fixer
-
-Fixes missing schema fields in ComfyUI workflow JSON files:
-- Adds missing 'flags', 'order', 'mode', 'properties', 'size' fields to nodes
-- Reconstructs 'inputs' and 'outputs' arrays from links
-- Builds complete 'links' array
-- Updates outdated node names
-
-Usage:
-    python3 fix_workflows.py <workflow_directory>
-"""
-
-import json
-import sys
-from pathlib import Path
-from typing import Dict, List, Any
-
-# Node name mapping (old → new)
-NODE_NAME_MAPPING = {
-    'AnimateDiffLoader': 'AnimateDiffLoaderV1',
-    'VHSVideoCombine': 'VHS_VideoCombine',
-    'PreviewVideo': None,  # Remove - use VHS_VideoCombine with preview enabled
-    'SaveVideo': None,  # Remove - use VHS_VideoCombine
-    'IPAdapterApply': 'IPAdapter',
-    'IPAdapterApplyFace': 'IPAdapterFaceID',
-    'AudioSave': 'SaveAudio',
-    'AnimateDiffSampler': 'KSamplerAdvanced',  # AnimateDiff uses standard KSampler
-    'ADE_AnimateDiffSampler': 'KSamplerAdvanced',
-    'SeedGenerator': 'ImpactInt',  # Use Impact Pack integer node for seed generation
-    'BatchKSampler': 'KSampler',  # Standard KSampler can handle batches
-    'ImageBatchToList': 'GetImageSize',  # Placeholder - may need manual adjustment
-}
-
-# Default node sizes by category
-NODE_SIZES = {
-    'Loader': {'0': 350, '1': 100},
-    'Sampler': {'0': 315, '1': 474},
-    'Encoder': {'0': 400, '1': 200},
-    'Default': {'0': 315, '1': 100},
-}
-
-def get_node_size(node_type: str) -> Dict[str, int]:
-    """Get appropriate size for node based on type"""
-    if 'Loader' in node_type or 'Load' in node_type:
-        return NODE_SIZES['Loader']
-    elif 'Sampler' in node_type or 'KSampler' in node_type:
-        return NODE_SIZES['Sampler']
-    elif 'Encode' in node_type or 'CLIP' in node_type:
-        return NODE_SIZES['Encoder']
-    else:
-        return NODE_SIZES['Default']
-
-def fix_workflow(workflow_path: Path) -> bool:
-    """Fix a single workflow file"""
-    print(f"\n{'='*60}")
-    print(f"Processing: {workflow_path.name}")
-    print(f"{'='*60}")
-
-    try:
-        with open(workflow_path, 'r') as f:
-            workflow = json.load(f)
-    except json.JSONDecodeError as e:
-        print(f"✗ ERROR: Invalid JSON - {e}")
-        return False
-
-    if 'nodes' not in workflow:
-        print(f"✗ ERROR: No 'nodes' key in workflow")
-        return False
-
-    nodes = workflow['nodes']
-    links = workflow.get('links', [])
-
-    # Track changes
-    changes = {
-        'added_flags': 0,
-        'added_order': 0,
-        'added_mode': 0,
-        'added_properties': 0,
-        'added_size': 0,
-        'added_inputs': 0,
-        'added_outputs': 0,
-        'updated_node_names': 0,
-        'removed_nodes': 0,
-        'added_last_link_id': 0,
-        'added_links': 0,
-    }
-
-    # Build link index for quick lookup
-    link_index = {}
-    for link in links:
-        if len(link) >= 6:
-            link_id, src_node_id, src_slot, tgt_node_id, tgt_slot, data_type = link[:6]
-            link_index[link_id] = {
-                'source': {'node_id': src_node_id, 'slot': src_slot},
-                'target': {'node_id': tgt_node_id, 'slot': tgt_slot},
-                'type': data_type
-            }
-
-    # Build node ID index
-    node_by_id = {node['id']: node for node in nodes}
-
-    # Process each node
-    nodes_to_remove = []
-    for i, node in enumerate(nodes):
-        node_id = node.get('id')
-        node_type = node.get('type', '')
-
-        # Update node name if needed
-        if node_type in NODE_NAME_MAPPING:
-            new_name = NODE_NAME_MAPPING[node_type]
-            if new_name is None:
-                # Mark for removal
-                nodes_to_remove.append(i)
-                changes['removed_nodes'] += 1
-                print(f"  Removing deprecated node {node_id}: {node_type}")
-                continue
-            else:
-                print(f"  Updating node {node_id}: {node_type} → {new_name}")
-                node['type'] = new_name
-                node_type = new_name
-                changes['updated_node_names'] += 1
-
-        # Add missing flags
-        if 'flags' not in node:
-            node['flags'] = {}
-            changes['added_flags'] += 1
-
-        # Add missing order (will recalculate later based on dependencies)
-        if 'order' not in node:
-            node['order'] = i  # Temporary order
-            changes['added_order'] += 1
-
-        # Add missing mode (0 = execute, 4 = bypass)
-        if 'mode' not in node:
-            node['mode'] = 0
-            changes['added_mode'] += 1
-
-        # Add missing properties
-        if 'properties' not in node:
-            node['properties'] = {"Node name for S&R": node_type}
-            changes['added_properties'] += 1
-
-        # Add missing size
-        if 'size' not in node:
-            node['size'] = get_node_size(node_type)
-            changes['added_size'] += 1
-
-        # Reconstruct inputs from links
-        if 'inputs' not in node or not node['inputs']:
-            node_inputs = []
-            for link_id, link_data in link_index.items():
-                if link_data['target']['node_id'] == node_id:
-                    # This link targets this node
-                    # We need to know the input name, but we don't have it
-                    # For now, create a placeholder
-                    node_inputs.append({
-                        'name': f'input_{link_data["target"]["slot"]}',
-                        'type': link_data['type'],
-                        'link': link_id
-                    })
-            if node_inputs:
-                node['inputs'] = node_inputs
-                changes['added_inputs'] += 1
-
-        # Reconstruct outputs from links
-        if 'outputs' not in node or not node['outputs']:
-            node_outputs = {}
-            for link_id, link_data in link_index.items():
-                if link_data['source']['node_id'] == node_id:
-                    slot = link_data['source']['slot']
-                    if slot not in node_outputs:
-                        node_outputs[slot] = {
-                            'name': f'output_{slot}',
-                            'type': link_data['type'],
-                            'links': [],
-                            'slot_index': slot
-                        }
-                    node_outputs[slot]['links'].append(link_id)
-
-            if node_outputs:
-                node['outputs'] = list(node_outputs.values())
-                changes['added_outputs'] += 1
-
-    # Remove deprecated nodes
-    for i in reversed(nodes_to_remove):
-        del nodes[i]
-
-    # Recalculate execution order based on dependencies
-    if changes['added_order'] > 0 or changes['removed_nodes'] > 0:
-        calculate_execution_order(nodes, link_index)
-
-    # Add missing links array
-    if 'links' not in workflow:
-        workflow['links'] = []
-        changes['added_links'] = 1
-
-    # Add missing last_link_id
-    if 'last_link_id' not in workflow:
-        # Calculate from existing links
-        max_link_id = 0
-        if workflow.get('links'):
-            for link in workflow['links']:
-                if link and len(link) > 0:
-                    max_link_id = max(max_link_id, link[0])
-        workflow['last_link_id'] = max_link_id
-        changes['added_last_link_id'] = 1
-
-    # Update workflow
-    workflow['nodes'] = nodes
-
-    # Print summary
-    print(f"\nChanges made:")
-    for key, value in changes.items():
-        if value > 0:
-            print(f"  • {key.replace('_', ' ').title()}: {value}")
-
-    total_changes = sum(changes.values())
-    if total_changes == 0:
-        print(f"  ✓ No changes needed - workflow already valid")
-        return True
-
-    # Save fixed workflow
-    try:
-        with open(workflow_path, 'w') as f:
-            json.dump(workflow, f, indent=2)
-        print(f"\n✓ Successfully fixed and saved workflow")
-        return True
-    except Exception as e:
-        print(f"\n✗ ERROR saving workflow: {e}")
-        return False
-
-def calculate_execution_order(nodes: List[Dict], link_index: Dict):
-    """Calculate execution order based on node dependencies"""
-    # Build dependency graph
-    dependencies = {}
-    node_by_id = {node['id']: node for node in nodes}
-
-    for node in nodes:
-        node_id = node['id']
-        dependencies[node_id] = set()
-
-        # Find all nodes this node depends on (inputs)
-        for link_id, link_data in link_index.items():
-            if link_data['target']['node_id'] == node_id:
-                # This node depends on the source node
-                dependencies[node_id].add(link_data['source']['node_id'])
-
-    # Topological sort to determine execution order
-    visited = set()
-    order_counter = [0]
-
-    def visit(node_id):
-        if node_id in visited:
-            return
-        visited.add(node_id)
-
-        # Visit dependencies first
-        for dep_id in dependencies.get(node_id, []):
-            if dep_id in node_by_id:  # Skip if dependency not in current nodes
-                visit(dep_id)
-
-        # Assign order
-        if node_id in node_by_id:
-            node_by_id[node_id]['order'] = order_counter[0]
-            order_counter[0] += 1
-
-    # Visit all nodes
-    for node_id in node_by_id.keys():
-        visit(node_id)
-
-def main():
-    if len(sys.argv) < 2:
-        print("Usage: python3 fix_workflows.py <workflow_directory>")
-        sys.exit(1)
-
-    workflow_dir = Path(sys.argv[1])
-
-    if not workflow_dir.exists():
-        print(f"Error: Directory {workflow_dir} does not exist")
-        sys.exit(1)
-
-    # Find all JSON files recursively
-    workflow_files = list(workflow_dir.rglob('*.json'))
-
-    if not workflow_files:
-        print(f"No workflow JSON files found in {workflow_dir}")
-        sys.exit(1)
-
-    print(f"\nFound {len(workflow_files)} workflow files")
-
-    # Process each workflow
-    success_count = 0
-    for workflow_path in sorted(workflow_files):
-        if fix_workflow(workflow_path):
-            success_count += 1
-
-    # Summary
-    print(f"\n{'='*60}")
-    print(f"SUMMARY")
-    print(f"{'='*60}")
-    print(f"Total workflows: {len(workflow_files)}")
-    print(f"Successfully fixed: {success_count}")
-    print(f"Failed: {len(workflow_files) - success_count}")
-    print(f"{'='*60}\n")
-
-if __name__ == '__main__':
-    main()

comfyui/patches/diffrhythm-llamaconfig-fix.patch

@@ -0,0 +1,56 @@
+diff --git a/__init__.py b/__init__.py
+index 1234567..abcdefg 100644
+--- a/__init__.py
++++ b/__init__.py
+@@ -1,3 +1,51 @@
++"""
++DiffRhythm ComfyUI Node with LlamaConfig Patch
++
++PATCH: Fixes "The size of tensor a (32) must match the size of tensor b (64)" error
++in DiffRhythm's rotary position embeddings by patching LlamaConfig initialization.
++
++Issue: DiffRhythm's DIT model doesn't specify num_attention_heads and
++num_key_value_heads when creating LlamaConfig, causing transformers 4.49.0+
++to incorrectly infer head_dim = 32 instead of 64.
++
++Solution: Patch LlamaConfig globally before importing DiffRhythmNode.
++
++Reference: https://github.com/billwuhao/ComfyUI_DiffRhythm/issues/44
++Reference: https://github.com/billwuhao/ComfyUI_DiffRhythm/issues/48
++
++Patch author: valknar@pivoine.art
++"""
++
++# CRITICAL: Patch LlamaConfig BEFORE importing DiffRhythmNode
++from transformers.models.llama import LlamaConfig as _OriginalLlamaConfig
++
++class PatchedLlamaConfig(_OriginalLlamaConfig):
++    """
++    Patched LlamaConfig that automatically adds missing attention head parameters.
++
++    Standard Llama architecture assumptions:
++    - head_dim = 64 (fixed)
++    - num_attention_heads = hidden_size // head_dim
++    - num_key_value_heads = num_attention_heads // 4 (for GQA)
++    """
++    def __init__(self, *args, **kwargs):
++        # If hidden_size is provided but num_attention_heads is not, calculate it
++        if 'hidden_size' in kwargs and 'num_attention_heads' not in kwargs:
++            hidden_size = kwargs['hidden_size']
++            kwargs['num_attention_heads'] = hidden_size // 64
++
++        # If num_key_value_heads is not provided, use GQA configuration
++        if 'num_attention_heads' in kwargs and 'num_key_value_heads' not in kwargs:
++            kwargs['num_key_value_heads'] = max(1, kwargs['num_attention_heads'] // 4)
++
++        super().__init__(*args, **kwargs)
++
++# Replace LlamaConfig in transformers module BEFORE DiffRhythm imports it
++import transformers.models.llama
++transformers.models.llama.LlamaConfig = PatchedLlamaConfig
++import transformers.models.llama.modeling_llama
++transformers.models.llama.modeling_llama.LlamaConfig = PatchedLlamaConfig
++
+ from .DiffRhythmNode import  NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS
+
+ __all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS"]

comfyui/requirements.txt

@@ -1,7 +1,7 @@
 torch
 torchvision
 torchaudio
-transformers
+transformers==4.49.0
 diffusers>=0.31.0
 accelerate
 safetensors
@@ -19,3 +19,4 @@ insightface
 onnxruntime
 pyyaml
 imageio-ffmpeg
+torchcodec

comfyui/start.sh

@@ -16,20 +16,22 @@ export TQDM_DISABLE=1
 # Navigate to ComfyUI directory
 cd "${COMFYUI_DIR}" || exit 1
 
-# Activate virtual environment
+# Determine which Python to use
-if [ -d "venv" ]; then
+if [ -f "venv/bin/python" ]; then
-  echo "Activating ComfyUI virtual environment..."
+  PYTHON_BIN="venv/bin/python"
-  source venv/bin/activate
+  echo "Using ComfyUI virtual environment Python..."
 else
+  PYTHON_BIN="python3"
   echo "WARNING: venv not found, using system Python"
 fi
 
 echo "Starting ComfyUI on port 8188..."
 echo "Access at: http://localhost:8188"
 echo "Using HuggingFace cache: ${HF_CACHE}"
+echo "Python: ${PYTHON_BIN}"
 
 # Start ComfyUI with GPU support
-python main.py \
+exec "${PYTHON_BIN}" main.py \
   --listen 0.0.0.0 \
   --port 8188 \
   --enable-cors-header \

comfyui/workflows/image-to-image/ipadapter-style-i2i-production-v1.json

@@ -47,7 +47,7 @@
         "Node name for S&R": "CheckpointLoaderSimple"
       },
       "widgets_values": [
-        "diffusers/stable-diffusion-xl-base-1.0"
+        "sd_xl_base_1.0.safetensors"
       ],
       "title": "SDXL Base Checkpoint Loader"
     },
@@ -140,7 +140,7 @@
         "Node name for S&R": "IPAdapterUnifiedLoader"
       },
       "widgets_values": [
-        "PLUS (high strength)"
+        "VIT-G (medium strength)"
       ],
       "title": "IP-Adapter Loader"
     },
@@ -197,10 +197,8 @@
       "widgets_values": [
         0.75,
         0.0,
-        "original",
-        0.0,
         1.0,
-        false
+        "style transfer"
       ],
       "title": "Apply IP-Adapter Style"
     },

comfyui/workflows/image-to-video/cogvideox-i2v-production-v1.json

@@ -1,201 +1,420 @@
 {
-  "last_node_id": 10,
+  "last_node_id": 8,
-  "last_link_id": 12,
+  "last_link_id": 10,
   "nodes": [
     {
       "id": 1,
       "type": "LoadImage",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [315, 314],
-        100
+      "widgets_values": ["input_frame.png", "image"],
-      ],
-      "size": [
-        315,
-        314
-      ],
-      "widgets_values": [
-        "input_frame.png",
-        "image"
-      ],
       "title": "API Input Image",
       "flags": {},
       "order": 0,
       "mode": 0,
       "properties": {
         "Node name for S&R": "LoadImage"
-      }
+      },
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null
+        }
+      ]
     },
     {
       "id": 2,
-      "type": "DiffusersLoader",
+      "type": "DownloadAndLoadCogVideoModel",
-      "pos": [
+      "pos": [50, 500],
-        50,
+      "size": [350, 100],
-        500
+      "widgets_values": ["THUDM/CogVideoX-5b-I2V", "bf16", "disabled", true],
-      ],
+      "title": "CogVideoX-5b-I2V Loader",
-      "widgets_values": [
-        "diffusion_models/CogVideoX-5b"
-      ],
-      "title": "CogVideoX-5b Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "DiffusersLoader"
+        "Node name for S&R": "DownloadAndLoadCogVideoModel"
       },
-      "size": {
+      "outputs": [
-        "0": 350,
+        {
-        "1": 100
+          "name": "model",
-      }
+          "type": "COGVIDEOMODEL",
+          "links": [2],
+          "slot_index": 0
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "links": [3, 10],
+          "slot_index": 1
+        }
+      ]
     },
     {
-      "id": 3,
+      "id": 7,
-      "type": "CLIPTextEncode",
+      "type": "CLIPLoader",
-      "pos": [
+      "pos": [50, 650],
-        450,
+      "size": [350, 100],
-        100
+      "widgets_values": ["t5xxl_fp16.safetensors", "sd3"],
-      ],
+      "title": "T5 CLIP Loader",
-      "widgets_values": [
-        "Camera movement description, action, scene details"
-      ],
-      "title": "API Video Prompt",
       "flags": {},
       "order": 2,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "CLIPTextEncode"
+        "Node name for S&R": "CLIPLoader"
       },
-      "size": {
+      "outputs": [
-        "0": 400,
+        {
-        "1": 200
+          "name": "CLIP",
-      }
+          "type": "CLIP",
+          "links": [4, 5],
+          "slot_index": 0
+        }
+      ]
     },
     {
-      "id": 4,
+      "id": 3,
-      "type": "CogVideoXSampler",
+      "type": "CogVideoTextEncode",
-      "pos": [
+      "pos": [450, 100],
-        800,
+      "size": [400, 200],
-        100
+      "widgets_values": ["Camera movement description, action, scene details", 1, false],
-      ],
+      "title": "API Video Prompt (Positive)",
-      "widgets_values": [
-        42,
-        "fixed",
-        50,
-        6.0,
-        49,
-        6
-      ],
-      "title": "CogVideoX Sampler (6s @ 8fps)",
       "flags": {},
       "order": 3,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "CogVideoXSampler"
+        "Node name for S&R": "CogVideoTextEncode"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 474
+          "name": "clip",
-      }
+          "type": "CLIP",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "conditioning",
+          "type": "CONDITIONING",
+          "links": [6],
+          "slot_index": 0
+        },
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "links": null
+        }
+      ]
     },
     {
-      "id": 5,
+      "id": 8,
-      "type": "VAEDecode",
+      "type": "CogVideoTextEncode",
-      "pos": [
+      "pos": [450, 350],
-        1150,
+      "size": [400, 200],
-        100
+      "widgets_values": ["low quality, blurry, distorted, watermark", 1, true],
-      ],
+      "title": "API Video Prompt (Negative)",
-      "title": "VAE Decode Video",
       "flags": {},
       "order": 4,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "VAEDecode"
+        "Node name for S&R": "CogVideoTextEncode"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 100
+          "name": "clip",
-      }
+          "type": "CLIP",
+          "link": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "conditioning",
+          "type": "CONDITIONING",
+          "links": [7],
+          "slot_index": 0
+        },
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "links": null
+        }
+      ]
     },
     {
-      "id": 6,
+      "id": 9,
-      "type": "VHS_VideoCombine",
+      "type": "CogVideoImageEncode",
-      "pos": [
+      "pos": [450, 600],
-        1450,
+      "size": [315, 100],
-        100
+      "widgets_values": [],
-      ],
+      "title": "Encode Input Image",
-      "widgets_values": [
-        8,
-        0,
-        "cogvideox_output",
-        "video/h264-mp4"
-      ],
-      "title": "Combine Video Frames",
       "flags": {},
       "order": 5,
       "mode": 0,
+      "properties": {
+        "Node name for S&R": "CogVideoImageEncode"
+      },
+      "inputs": [
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 3
+        },
+        {
+          "name": "start_image",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "links": [8],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 4,
+      "type": "CogVideoSampler",
+      "pos": [900, 100],
+      "size": [315, 474],
+      "widgets_values": [49, 50, 6.0, 42, "fixed", "CogVideoXDDIM", 1.0],
+      "title": "CogVideoX Sampler (6s @ 8fps)",
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "properties": {
+        "Node name for S&R": "CogVideoSampler"
+      },
+      "inputs": [
+        {
+          "name": "model",
+          "type": "COGVIDEOMODEL",
+          "link": 2
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 6
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 7
+        },
+        {
+          "name": "image_cond_latents",
+          "type": "LATENT",
+          "link": 8
+        }
+      ],
+      "outputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "links": [9],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 5,
+      "type": "CogVideoDecode",
+      "pos": [1250, 100],
+      "size": [315, 200],
+      "widgets_values": [true, 240, 360, 0.25, 0.25],
+      "title": "VAE Decode Video",
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "properties": {
+        "Node name for S&R": "CogVideoDecode"
+      },
+      "inputs": [
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 10
+        },
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 9
+        }
+      ],
+      "outputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "links": [11],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 6,
+      "type": "VHS_VideoCombine",
+      "pos": [1600, 100],
+      "size": [315, 200],
+      "widgets_values": [8, 0, "cogvideox_output", "video/h264-mp4", "yuv420p", 19, true, false],
+      "title": "Combine Video Frames",
+      "flags": {},
+      "order": 8,
+      "mode": 0,
       "properties": {
         "Node name for S&R": "VHS_VideoCombine"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 100
+          "name": "images",
-      }
+          "type": "IMAGE",
+          "link": 11
+        }
+      ],
+      "outputs": [
+        {
+          "name": "Filenames",
+          "type": "VHS_FILENAMES",
+          "links": null
+        }
+      ]
     }
   ],
-  "links": [],
+  "links": [
+    [1, 1, 0, 9, 1, "IMAGE"],
+    [2, 2, 0, 4, 0, "COGVIDEOMODEL"],
+    [3, 2, 1, 9, 0, "VAE"],
+    [4, 7, 0, 3, 0, "CLIP"],
+    [5, 7, 0, 8, 0, "CLIP"],
+    [6, 3, 0, 4, 1, "CONDITIONING"],
+    [7, 8, 0, 4, 2, "CONDITIONING"],
+    [8, 9, 0, 4, 3, "LATENT"],
+    [9, 4, 0, 5, 1, "LATENT"],
+    [10, 2, 1, 5, 0, "VAE"],
+    [11, 5, 0, 6, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
       "name": "CogVideoX Image-to-Video Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
-      "description": "AI-driven image-to-video using CogVideoX-5b. Generate 6-second videos (48 frames @ 8fps) from input images with camera movement and action.",
+      "description": "AI-driven image-to-video using CogVideoX-5b-I2V. Generate 6-second videos (49 frames @ 8fps) from input images with camera movement and action.",
       "category": "image-to-video",
-      "tags": [
+      "tags": ["cogvideox", "i2v", "video-generation", "production"],
-        "cogvideox",
-        "i2v",
-        "video-generation",
-        "production"
-      ],
       "requirements": {
-        "models": [
+        "models": ["CogVideoX-5b-I2V", "T5-XXL FP16"],
-          "CogVideoX-5b"
+        "custom_nodes": ["ComfyUI-VideoHelperSuite", "ComfyUI-CogVideoXWrapper"],
-        ],
+        "vram_min": "20GB",
-        "custom_nodes": [
+        "vram_recommended": "24GB"
-          "ComfyUI-VideoHelperSuite",
-          "ComfyUI-CogVideoXWrapper"
-        ],
-        "vram_min": "20GB"
       },
       "parameters": {
         "input_image": {
           "node_id": 1,
+          "widget_index": 0,
           "type": "image",
           "required": true,
-          "description": "Starting frame for video"
+          "description": "Starting frame for video generation"
         },
-        "video_prompt": {
+        "positive_prompt": {
           "node_id": 3,
+          "widget_index": 0,
           "type": "string",
           "required": true,
-          "description": "Describe camera movement and action"
+          "default": "Camera movement description, action, scene details",
+          "description": "Describe desired camera movement, actions, and scene"
+        },
+        "negative_prompt": {
+          "node_id": 8,
+          "widget_index": 0,
+          "type": "string",
+          "required": false,
+          "default": "low quality, blurry, distorted, watermark",
+          "description": "Undesired elements to avoid"
+        },
+        "num_frames": {
+          "node_id": 4,
+          "widget_index": 0,
+          "type": "integer",
+          "required": false,
+          "default": 49,
+          "min": 1,
+          "max": 1024,
+          "description": "Number of frames to generate (49 = ~6s @ 8fps)"
         },
         "steps": {
           "node_id": 4,
+          "widget_index": 1,
           "type": "integer",
+          "required": false,
           "default": 50,
-          "description": "Sampling steps (50 recommended)"
+          "min": 20,
+          "max": 100,
+          "description": "Sampling steps (50 recommended for quality)"
+        },
+        "cfg": {
+          "node_id": 4,
+          "widget_index": 2,
+          "type": "float",
+          "required": false,
+          "default": 6.0,
+          "min": 1.0,
+          "max": 30.0,
+          "description": "Classifier-free guidance scale"
+        },
+        "seed": {
+          "node_id": 4,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 42,
+          "description": "Random seed for reproducibility"
         },
         "fps": {
           "node_id": 6,
+          "widget_index": 0,
           "type": "integer",
+          "required": false,
           "default": 8,
-          "description": "Output framerate"
+          "description": "Output video framerate"
+        }
+      },
+      "outputs": {
+        "video": {
+          "node_id": 6,
+          "type": "video",
+          "format": "MP4 (H.264)",
+          "resolution": "Based on input image",
+          "duration": "~6 seconds @ 8fps (49 frames)"
         }
       },
       "performance": {
         "avg_generation_time": "120-180 seconds",
         "vram_usage": "~20-22GB",
-        "output": "6 seconds @ 8fps (48 frames)"
+        "gpu_utilization": "95-100%"
-      }
+      },
+      "use_cases": [
+        "Animate static images with camera motion",
+        "Create video loops from single frames",
+        "Add dynamic movement to product shots",
+        "Generate cinematic camera movements"
+      ],
+      "notes": [
+        "CogVideoX-5b-I2V is specifically trained for image-to-video generation",
+        "Model will download automatically on first use (~10GB)",
+        "Enable VAE tiling to reduce VRAM usage",
+        "Higher steps (50-100) improve quality but increase generation time",
+        "T5-XXL text encoder required - automatically linked from SD3.5"
+      ]
     }
   },
   "version": 0.4
-}
+}

comfyui/workflows/image-to-video/i2v_wan22-5b-ti2v.json

@@ -0,0 +1,733 @@
+{
+  "id": "91f6bbe2-ed41-4fd6-bac7-71d5b5864ecb",
+  "revision": 0,
+  "last_node_id": 59,
+  "last_link_id": 108,
+  "nodes": [
+    {
+      "id": 37,
+      "type": "UNETLoader",
+      "pos": [
+        -30,
+        50
+      ],
+      "size": [
+        346.7470703125,
+        82
+      ],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "slot_index": 0,
+          "links": [
+            94
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "UNETLoader",
+        "models": [
+          {
+            "name": "wan2.2_ti2v_5B_fp16.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors",
+            "directory": "diffusion_models"
+          }
+        ]
+      },
+      "widgets_values": [
+        "wan2.2_ti2v_5B_fp16.safetensors",
+        "default"
+      ]
+    },
+    {
+      "id": 38,
+      "type": "CLIPLoader",
+      "pos": [
+        -30,
+        190
+      ],
+      "size": [
+        350,
+        110
+      ],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "slot_index": 0,
+          "links": [
+            74,
+            75
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "CLIPLoader",
+        "models": [
+          {
+            "name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors",
+            "directory": "text_encoders"
+          }
+        ]
+      },
+      "widgets_values": [
+        "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
+        "wan",
+        "default"
+      ]
+    },
+    {
+      "id": 39,
+      "type": "VAELoader",
+      "pos": [
+        -30,
+        350
+      ],
+      "size": [
+        350,
+        60
+      ],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "slot_index": 0,
+          "links": [
+            76,
+            105
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "VAELoader",
+        "models": [
+          {
+            "name": "wan2.2_vae.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors",
+            "directory": "vae"
+          }
+        ]
+      },
+      "widgets_values": [
+        "wan2.2_vae.safetensors"
+      ]
+    },
+    {
+      "id": 8,
+      "type": "VAEDecode",
+      "pos": [
+        1190,
+        150
+      ],
+      "size": [
+        210,
+        46
+      ],
+      "flags": {},
+      "order": 10,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 35
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 76
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "slot_index": 0,
+          "links": [
+            107
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "VAEDecode"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 57,
+      "type": "CreateVideo",
+      "pos": [
+        1200,
+        240
+      ],
+      "size": [
+        270,
+        78
+      ],
+      "flags": {},
+      "order": 11,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 107
+        },
+        {
+          "name": "audio",
+          "shape": 7,
+          "type": "AUDIO",
+          "link": null
+        }
+      ],
+      "outputs": [
+        {
+          "name": "VIDEO",
+          "type": "VIDEO",
+          "links": [
+            108
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "CreateVideo"
+      },
+      "widgets_values": [
+        24
+      ]
+    },
+    {
+      "id": 58,
+      "type": "SaveVideo",
+      "pos": [
+        1200,
+        370
+      ],
+      "size": [
+        660,
+        450
+      ],
+      "flags": {},
+      "order": 12,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "video",
+          "type": "VIDEO",
+          "link": 108
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "SaveVideo"
+      },
+      "widgets_values": [
+        "video/ComfyUI",
+        "auto",
+        "auto"
+      ]
+    },
+    {
+      "id": 55,
+      "type": "Wan22ImageToVideoLatent",
+      "pos": [
+        380,
+        540
+      ],
+      "size": [
+        271.9126892089844,
+        150
+      ],
+      "flags": {},
+      "order": 8,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 105
+        },
+        {
+          "name": "start_image",
+          "shape": 7,
+          "type": "IMAGE",
+          "link": 106
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            104
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "Wan22ImageToVideoLatent"
+      },
+      "widgets_values": [
+        1280,
+        704,
+        121,
+        1
+      ]
+    },
+    {
+      "id": 56,
+      "type": "LoadImage",
+      "pos": [
+        0,
+        540
+      ],
+      "size": [
+        274.080078125,
+        314
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 4,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            106
+          ]
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "LoadImage"
+      },
+      "widgets_values": [
+        "example.png",
+        "image"
+      ]
+    },
+    {
+      "id": 7,
+      "type": "CLIPTextEncode",
+      "pos": [
+        380,
+        260
+      ],
+      "size": [
+        425.27801513671875,
+        180.6060791015625
+      ],
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 75
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "slot_index": 0,
+          "links": [
+            52
+          ]
+        }
+      ],
+      "title": "CLIP Text Encode (Negative Prompt)",
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": [
+        "色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走"
+      ],
+      "color": "#322",
+      "bgcolor": "#533"
+    },
+    {
+      "id": 6,
+      "type": "CLIPTextEncode",
+      "pos": [
+        380,
+        50
+      ],
+      "size": [
+        422.84503173828125,
+        164.31304931640625
+      ],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 74
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "slot_index": 0,
+          "links": [
+            46
+          ]
+        }
+      ],
+      "title": "CLIP Text Encode (Positive Prompt)",
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": [
+        "Low contrast. In a retro 1970s-style subway station, a street musician plays in dim colors and rough textures. He wears an old jacket, playing guitar with focus. Commuters hurry by, and a small crowd gathers to listen. The camera slowly moves right, capturing the blend of music and city noise, with old subway signs and mottled walls in the background."
+      ],
+      "color": "#232",
+      "bgcolor": "#353"
+    },
+    {
+      "id": 3,
+      "type": "KSampler",
+      "pos": [
+        850,
+        130
+      ],
+      "size": [
+        315,
+        262
+      ],
+      "flags": {},
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 95
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 46
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 52
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 104
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "slot_index": 0,
+          "links": [
+            35
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        898471028164125,
+        "randomize",
+        20,
+        5,
+        "uni_pc",
+        "simple",
+        1
+      ]
+    },
+    {
+      "id": 48,
+      "type": "ModelSamplingSD3",
+      "pos": [
+        850,
+        20
+      ],
+      "size": [
+        210,
+        58
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 94
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "slot_index": 0,
+          "links": [
+            95
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.45",
+        "Node name for S&R": "ModelSamplingSD3"
+      },
+      "widgets_values": [
+        8
+      ]
+    },
+    {
+      "id": 59,
+      "type": "MarkdownNote",
+      "pos": [
+        -550,
+        10
+      ],
+      "size": [
+        480,
+        340
+      ],
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "title": "Model Links",
+      "properties": {},
+      "widgets_values": [
+        "[Tutorial](https://docs.comfy.org/tutorials/video/wan/wan2_2\n) \n\n**Diffusion Model**\n- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors)\n\n**VAE**\n- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors)\n\n**Text Encoder**   \n- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors)\n\n\nFile save location\n\n```\nComfyUI/\n├───📂 models/\n│   ├───📂 diffusion_models/\n│   │   └───wan2.2_ti2v_5B_fp16.safetensors\n│   ├───📂 text_encoders/\n│   │   └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors \n│   └───📂 vae/\n│       └── wan2.2_vae.safetensors\n```\n"
+      ],
+      "color": "#432",
+      "bgcolor": "#653"
+    }
+  ],
+  "links": [
+    [
+      35,
+      3,
+      0,
+      8,
+      0,
+      "LATENT"
+    ],
+    [
+      46,
+      6,
+      0,
+      3,
+      1,
+      "CONDITIONING"
+    ],
+    [
+      52,
+      7,
+      0,
+      3,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      74,
+      38,
+      0,
+      6,
+      0,
+      "CLIP"
+    ],
+    [
+      75,
+      38,
+      0,
+      7,
+      0,
+      "CLIP"
+    ],
+    [
+      76,
+      39,
+      0,
+      8,
+      1,
+      "VAE"
+    ],
+    [
+      94,
+      37,
+      0,
+      48,
+      0,
+      "MODEL"
+    ],
+    [
+      95,
+      48,
+      0,
+      3,
+      0,
+      "MODEL"
+    ],
+    [
+      104,
+      55,
+      0,
+      3,
+      3,
+      "LATENT"
+    ],
+    [
+      105,
+      39,
+      0,
+      55,
+      0,
+      "VAE"
+    ],
+    [
+      106,
+      56,
+      0,
+      55,
+      1,
+      "IMAGE"
+    ],
+    [
+      107,
+      8,
+      0,
+      57,
+      0,
+      "IMAGE"
+    ],
+    [
+      108,
+      57,
+      0,
+      58,
+      0,
+      "VIDEO"
+    ]
+  ],
+  "groups": [
+    {
+      "id": 1,
+      "title": "Step1 - Load models",
+      "bounding": [
+        -50,
+        -20,
+        400,
+        453.6000061035156
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 2,
+      "title": "Step3 - Prompt",
+      "bounding": [
+        370,
+        -20,
+        448.27801513671875,
+        473.2060852050781
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 3,
+      "title": "For i2v, use Ctrl + B to enable",
+      "bounding": [
+        -50,
+        450,
+        400,
+        420
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 4,
+      "title": "Video Size & length",
+      "bounding": [
+        370,
+        470,
+        291.9127197265625,
+        233.60000610351562
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    }
+  ],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 0.46462425349300085,
+      "offset": [
+        847.5372059811432,
+        288.7938392118285
+      ]
+    },
+    "frontendVersion": "1.27.10",
+    "VHS_latentpreview": false,
+    "VHS_latentpreviewrate": 0,
+    "VHS_MetadataImage": true,
+    "VHS_KeepIntermediate": true
+  },
+  "version": 0.4
+}

comfyui/workflows/image-to-video/svd-i2v-production-v1.json

@@ -1,18 +1,13 @@
 {
-  "last_node_id": 8,
+  "last_node_id": 7,
   "last_link_id": 10,
   "nodes": [
     {
       "id": 1,
       "type": "LoadImage",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [315, 314],
-        100
+      "widgets_values": ["input_frame.png", "image"],
-      ],
-      "widgets_values": [
-        "input_frame.png",
-        "image"
-      ],
       "title": "API Input Image",
       "flags": {},
       "order": 0,
@@ -20,158 +15,402 @@
       "properties": {
         "Node name for S&R": "LoadImage"
       },
-      "size": {
+      "outputs": [
-        "0": 350,
+        {
-        "1": 100
+          "name": "IMAGE",
-      }
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null
+        }
+      ]
     },
     {
       "id": 2,
-      "type": "DiffusersLoader",
+      "type": "ImageOnlyCheckpointLoader",
-      "pos": [
+      "pos": [50, 500],
-        50,
+      "size": [350, 100],
-        400
+      "widgets_values": ["svd_xt.safetensors"],
-      ],
+      "title": "SVD-XT Model Loader",
-      "widgets_values": [
-        "diffusion_models/stable-video-diffusion-img2vid"
-      ],
-      "title": "SVD Model Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "DiffusersLoader"
+        "Node name for S&R": "ImageOnlyCheckpointLoader"
       },
-      "size": {
+      "outputs": [
-        "0": 350,
+        {
-        "1": 100
+          "name": "MODEL",
-      }
+          "type": "MODEL",
+          "links": [2],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP_VISION",
+          "type": "CLIP_VISION",
+          "links": [3],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [4, 5],
+          "slot_index": 2
+        }
+      ]
     },
     {
       "id": 3,
-      "type": "SVDSampler",
+      "type": "VideoLinearCFGGuidance",
-      "pos": [
+      "pos": [450, 500],
-        450,
+      "size": [315, 100],
-        100
+      "widgets_values": [1.0],
-      ],
+      "title": "Linear CFG Guidance",
-      "widgets_values": [
-        42,
-        "fixed",
-        25,
-        14,
-        127,
-        0.02
-      ],
-      "title": "SVD Sampler (14 frames)",
       "flags": {},
       "order": 2,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "SVDSampler"
+        "Node name for S&R": "VideoLinearCFGGuidance"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 474
+          "name": "model",
-      }
+          "type": "MODEL",
+          "link": 2
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [6],
+          "slot_index": 0
+        }
+      ]
     },
     {
       "id": 4,
-      "type": "VAEDecode",
+      "type": "SVD_img2vid_Conditioning",
-      "pos": [
+      "pos": [450, 100],
-        800,
+      "size": [315, 350],
-        100
+      "widgets_values": [1024, 576, 14, 127, 6, 0.0],
-      ],
+      "title": "SVD Image-to-Video Conditioning",
-      "title": "VAE Decode Video",
       "flags": {},
       "order": 3,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "VAEDecode"
+        "Node name for S&R": "SVD_img2vid_Conditioning"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 100
+          "name": "clip_vision",
-      }
+          "type": "CLIP_VISION",
+          "link": 3
+        },
+        {
+          "name": "init_image",
+          "type": "IMAGE",
+          "link": 1
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "links": [7],
+          "slot_index": 0
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "links": [8],
+          "slot_index": 1
+        },
+        {
+          "name": "latent",
+          "type": "LATENT",
+          "links": [9],
+          "slot_index": 2
+        }
+      ]
     },
     {
       "id": 5,
-      "type": "VHS_VideoCombine",
+      "type": "KSampler",
-      "pos": [
+      "pos": [800, 100],
-        1100,
+      "size": [315, 474],
-        100
+      "widgets_values": [42, "fixed", 25, 6.0, "euler", "karras", 1.0],
-      ],
+      "title": "KSampler (25 steps)",
-      "widgets_values": [
-        6,
-        0,
-        "svd_output",
-        "video/h264-mp4"
-      ],
-      "title": "Combine Frames",
       "flags": {},
       "order": 4,
       "mode": 0,
+      "properties": {
+        "Node name for S&R": "KSampler"
+      },
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 6
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 7
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 8
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 9
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [10],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 6,
+      "type": "VAEDecode",
+      "pos": [1150, 100],
+      "size": [210, 46],
+      "widgets_values": [],
+      "title": "VAE Decode Video Frames",
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 10
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [11],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 7,
+      "type": "VHS_VideoCombine",
+      "pos": [1400, 100],
+      "size": [315, 200],
+      "widgets_values": [6, 0, "svd_output", "video/h264-mp4", false, true],
+      "title": "Combine Video Frames",
+      "flags": {},
+      "order": 6,
+      "mode": 0,
       "properties": {
         "Node name for S&R": "VHS_VideoCombine"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 100
+          "name": "images",
-      }
+          "type": "IMAGE",
+          "link": 11
+        }
+      ],
+      "outputs": [
+        {
+          "name": "Filenames",
+          "type": "VHS_FILENAMES",
+          "links": null
+        }
+      ]
     }
   ],
-  "links": [],
+  "links": [
+    [1, 1, 0, 4, 1, "IMAGE"],
+    [2, 2, 0, 3, 0, "MODEL"],
+    [3, 2, 1, 4, 0, "CLIP_VISION"],
+    [4, 2, 2, 4, 2, "VAE"],
+    [5, 2, 2, 6, 1, "VAE"],
+    [6, 3, 0, 5, 0, "MODEL"],
+    [7, 4, 0, 5, 1, "CONDITIONING"],
+    [8, 4, 1, 5, 2, "CONDITIONING"],
+    [9, 4, 2, 5, 3, "LATENT"],
+    [10, 5, 0, 6, 0, "LATENT"],
+    [11, 6, 0, 7, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
       "name": "Stable Video Diffusion Image-to-Video Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
-      "description": "Quick animation using SVD. Generate 14-frame video from single image with motion and camera movement.",
+      "description": "Quick animation using SVD-XT. Generate 14-frame video from single image with motion and camera movement.",
       "category": "image-to-video",
-      "tags": [
+      "tags": ["svd", "svd-xt", "stable-video-diffusion", "i2v", "animation", "production"],
-        "svd",
-        "stable-video-diffusion",
-        "i2v",
-        "animation",
-        "production"
-      ],
       "requirements": {
-        "models": [
+        "models": ["SVD-XT"],
-          "stable-video-diffusion-img2vid"
+        "custom_nodes": ["ComfyUI-VideoHelperSuite"],
-        ],
+        "vram_min": "16GB",
-        "custom_nodes": [
+        "vram_recommended": "20GB"
-          "ComfyUI-VideoHelperSuite"
-        ],
-        "vram_min": "16GB"
       },
       "parameters": {
         "input_image": {
           "node_id": 1,
+          "widget_index": 0,
           "type": "image",
-          "required": true
+          "required": true,
+          "description": "Starting frame for video generation (1024x576 recommended)"
+        },
+        "width": {
+          "node_id": 4,
+          "widget_index": 0,
+          "type": "integer",
+          "required": false,
+          "default": 1024,
+          "min": 16,
+          "max": 16384,
+          "description": "Output video width"
+        },
+        "height": {
+          "node_id": 4,
+          "widget_index": 1,
+          "type": "integer",
+          "required": false,
+          "default": 576,
+          "min": 16,
+          "max": 16384,
+          "description": "Output video height"
+        },
+        "video_frames": {
+          "node_id": 4,
+          "widget_index": 2,
+          "type": "integer",
+          "required": false,
+          "default": 14,
+          "min": 1,
+          "max": 4096,
+          "description": "Number of frames to generate (14 or 25 for SVD/SVD-XT)"
+        },
+        "motion_bucket_id": {
+          "node_id": 4,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 127,
+          "min": 1,
+          "max": 1023,
+          "description": "Motion amount (higher = more motion)"
+        },
+        "fps": {
+          "node_id": 4,
+          "widget_index": 4,
+          "type": "integer",
+          "required": false,
+          "default": 6,
+          "min": 1,
+          "max": 1024,
+          "description": "Frames per second for conditioning"
+        },
+        "augmentation_level": {
+          "node_id": 4,
+          "widget_index": 5,
+          "type": "float",
+          "required": false,
+          "default": 0.0,
+          "min": 0.0,
+          "max": 10.0,
+          "description": "Noise augmentation level"
         },
         "steps": {
-          "node_id": 3,
+          "node_id": 5,
+          "widget_index": 2,
           "type": "integer",
-          "default": 25
+          "required": false,
+          "default": 25,
+          "min": 1,
+          "max": 150,
+          "description": "Sampling steps (25 recommended)"
         },
-        "frames": {
+        "cfg": {
-          "node_id": 3,
+          "node_id": 5,
-          "type": "integer",
+          "widget_index": 3,
-          "default": 14,
+          "type": "float",
-          "description": "Number of output frames"
+          "required": false,
+          "default": 6.0,
+          "min": 0.0,
+          "max": 30.0,
+          "description": "Classifier-free guidance scale"
         },
-        "motion_bucket": {
+        "seed": {
-          "node_id": 3,
+          "node_id": 5,
+          "widget_index": 0,
           "type": "integer",
-          "default": 127,
+          "required": false,
-          "description": "Motion amount (0-255)"
+          "default": 42,
+          "description": "Random seed for reproducibility"
+        },
+        "output_fps": {
+          "node_id": 7,
+          "widget_index": 0,
+          "type": "integer",
+          "required": false,
+          "default": 6,
+          "description": "Output video framerate"
+        }
+      },
+      "outputs": {
+        "video": {
+          "node_id": 7,
+          "type": "video",
+          "format": "MP4 (H.264)",
+          "resolution": "1024x576 (configurable)",
+          "duration": "~2.3 seconds @ 6fps (14 frames)"
         }
       },
       "performance": {
         "avg_generation_time": "25-35 seconds",
-        "vram_usage": "~14-16GB",
+        "vram_usage": "~16-18GB",
-        "output": "14 frames (~2.3s @ 6fps)"
+        "gpu_utilization": "95-100%"
-      }
+      },
+      "use_cases": [
+        "Animate static images with natural motion",
+        "Create short video loops from single frames",
+        "Add subtle camera movements to still images",
+        "Generate product animation previews"
+      ],
+      "notes": [
+        "SVD-XT extends frame count from 14 to 25 frames",
+        "Model auto-downloads on first use (~9GB)",
+        "Recommended resolution: 1024x576 (16:9)",
+        "Higher motion_bucket_id = more movement",
+        "Linear CFG guidance improves temporal consistency"
+      ]
     }
   },
   "version": 0.4
-}
+}

comfyui/workflows/image-to-video/svd-xt-i2v-production-v1.json

@@ -1,18 +1,13 @@
 {
-  "last_node_id": 8,
+  "last_node_id": 7,
   "last_link_id": 10,
   "nodes": [
     {
       "id": 1,
       "type": "LoadImage",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [315, 314],
-        100
+      "widgets_values": ["input_frame.png", "image"],
-      ],
-      "widgets_values": [
-        "input_frame.png",
-        "image"
-      ],
       "title": "API Input Image",
       "flags": {},
       "order": 0,
@@ -20,157 +15,403 @@
       "properties": {
         "Node name for S&R": "LoadImage"
       },
-      "size": {
+      "outputs": [
-        "0": 350,
+        {
-        "1": 100
+          "name": "IMAGE",
-      }
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null
+        }
+      ]
     },
     {
       "id": 2,
-      "type": "DiffusersLoader",
+      "type": "ImageOnlyCheckpointLoader",
-      "pos": [
+      "pos": [50, 500],
-        50,
+      "size": [350, 100],
-        400
+      "widgets_values": ["svd_xt.safetensors"],
-      ],
-      "widgets_values": [
-        "diffusion_models/stable-video-diffusion-img2vid-xt"
-      ],
       "title": "SVD-XT Model Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "DiffusersLoader"
+        "Node name for S&R": "ImageOnlyCheckpointLoader"
       },
-      "size": {
+      "outputs": [
-        "0": 350,
+        {
-        "1": 100
+          "name": "MODEL",
-      }
+          "type": "MODEL",
+          "links": [2],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP_VISION",
+          "type": "CLIP_VISION",
+          "links": [3],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [4, 5],
+          "slot_index": 2
+        }
+      ]
     },
     {
       "id": 3,
-      "type": "SVDSampler",
+      "type": "VideoLinearCFGGuidance",
-      "pos": [
+      "pos": [450, 500],
-        450,
+      "size": [315, 100],
-        100
+      "widgets_values": [1.0],
-      ],
+      "title": "Linear CFG Guidance",
-      "widgets_values": [
-        42,
-        "fixed",
-        30,
-        25,
-        127,
-        0.02
-      ],
-      "title": "SVD-XT Sampler (25 frames)",
       "flags": {},
       "order": 2,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "SVDSampler"
+        "Node name for S&R": "VideoLinearCFGGuidance"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 474
+          "name": "model",
-      }
+          "type": "MODEL",
+          "link": 2
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [6],
+          "slot_index": 0
+        }
+      ]
     },
     {
       "id": 4,
-      "type": "VAEDecode",
+      "type": "SVD_img2vid_Conditioning",
-      "pos": [
+      "pos": [450, 100],
-        800,
+      "size": [315, 350],
-        100
+      "widgets_values": [1024, 576, 25, 127, 6, 0.0],
-      ],
+      "title": "SVD-XT Image-to-Video Conditioning (25 frames)",
-      "title": "VAE Decode Video",
       "flags": {},
       "order": 3,
       "mode": 0,
       "properties": {
-        "Node name for S&R": "VAEDecode"
+        "Node name for S&R": "SVD_img2vid_Conditioning"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 100
+          "name": "clip_vision",
-      }
+          "type": "CLIP_VISION",
+          "link": 3
+        },
+        {
+          "name": "init_image",
+          "type": "IMAGE",
+          "link": 1
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "links": [7],
+          "slot_index": 0
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "links": [8],
+          "slot_index": 1
+        },
+        {
+          "name": "latent",
+          "type": "LATENT",
+          "links": [9],
+          "slot_index": 2
+        }
+      ]
     },
     {
       "id": 5,
-      "type": "VHS_VideoCombine",
+      "type": "KSampler",
-      "pos": [
+      "pos": [800, 100],
-        1100,
+      "size": [315, 474],
-        100
+      "widgets_values": [42, "fixed", 30, 6.0, "euler", "karras", 1.0],
-      ],
+      "title": "KSampler (30 steps)",
-      "widgets_values": [
-        6,
-        0,
-        "svd_xt_output",
-        "video/h264-mp4"
-      ],
-      "title": "Combine Frames",
       "flags": {},
       "order": 4,
       "mode": 0,
+      "properties": {
+        "Node name for S&R": "KSampler"
+      },
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 6
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 7
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 8
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 9
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [10],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 6,
+      "type": "VAEDecode",
+      "pos": [1150, 100],
+      "size": [210, 46],
+      "widgets_values": [],
+      "title": "VAE Decode Video Frames",
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 10
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [11],
+          "slot_index": 0
+        }
+      ]
+    },
+    {
+      "id": 7,
+      "type": "VHS_VideoCombine",
+      "pos": [1400, 100],
+      "size": [315, 200],
+      "widgets_values": [6, 0, "svd_xt_output", "video/h264-mp4", false, true],
+      "title": "Combine Video Frames",
+      "flags": {},
+      "order": 6,
+      "mode": 0,
       "properties": {
         "Node name for S&R": "VHS_VideoCombine"
       },
-      "size": {
+      "inputs": [
-        "0": 315,
+        {
-        "1": 100
+          "name": "images",
-      }
+          "type": "IMAGE",
+          "link": 11
+        }
+      ],
+      "outputs": [
+        {
+          "name": "Filenames",
+          "type": "VHS_FILENAMES",
+          "links": null
+        }
+      ]
     }
   ],
-  "links": [],
+  "links": [
+    [1, 1, 0, 4, 1, "IMAGE"],
+    [2, 2, 0, 3, 0, "MODEL"],
+    [3, 2, 1, 4, 0, "CLIP_VISION"],
+    [4, 2, 2, 4, 2, "VAE"],
+    [5, 2, 2, 6, 1, "VAE"],
+    [6, 3, 0, 5, 0, "MODEL"],
+    [7, 4, 0, 5, 1, "CONDITIONING"],
+    [8, 4, 1, 5, 2, "CONDITIONING"],
+    [9, 4, 2, 5, 3, "LATENT"],
+    [10, 5, 0, 6, 0, "LATENT"],
+    [11, 6, 0, 7, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
       "name": "Stable Video Diffusion XT Image-to-Video Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
       "description": "Extended animation using SVD-XT. Generate 25-frame video for longer animations with smooth motion.",
       "category": "image-to-video",
-      "tags": [
+      "tags": ["svd-xt", "stable-video-diffusion", "i2v", "extended", "production"],
-        "svd-xt",
-        "stable-video-diffusion",
-        "i2v",
-        "extended",
-        "production"
-      ],
       "requirements": {
-        "models": [
+        "models": ["SVD-XT"],
-          "stable-video-diffusion-img2vid-xt"
+        "custom_nodes": ["ComfyUI-VideoHelperSuite"],
-        ],
+        "vram_min": "18GB",
-        "custom_nodes": [
+        "vram_recommended": "20GB"
-          "ComfyUI-VideoHelperSuite"
-        ],
-        "vram_min": "18GB"
       },
       "parameters": {
         "input_image": {
           "node_id": 1,
+          "widget_index": 0,
           "type": "image",
-          "required": true
+          "required": true,
+          "description": "Starting frame for video generation (1024x576 recommended)"
+        },
+        "width": {
+          "node_id": 4,
+          "widget_index": 0,
+          "type": "integer",
+          "required": false,
+          "default": 1024,
+          "min": 16,
+          "max": 16384,
+          "description": "Output video width"
+        },
+        "height": {
+          "node_id": 4,
+          "widget_index": 1,
+          "type": "integer",
+          "required": false,
+          "default": 576,
+          "min": 16,
+          "max": 16384,
+          "description": "Output video height"
+        },
+        "video_frames": {
+          "node_id": 4,
+          "widget_index": 2,
+          "type": "integer",
+          "required": false,
+          "default": 25,
+          "min": 1,
+          "max": 4096,
+          "description": "Number of frames to generate (25 for SVD-XT)"
+        },
+        "motion_bucket_id": {
+          "node_id": 4,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 127,
+          "min": 1,
+          "max": 1023,
+          "description": "Motion amount (higher = more motion)"
+        },
+        "fps": {
+          "node_id": 4,
+          "widget_index": 4,
+          "type": "integer",
+          "required": false,
+          "default": 6,
+          "min": 1,
+          "max": 1024,
+          "description": "Frames per second for conditioning"
+        },
+        "augmentation_level": {
+          "node_id": 4,
+          "widget_index": 5,
+          "type": "float",
+          "required": false,
+          "default": 0.0,
+          "min": 0.0,
+          "max": 10.0,
+          "description": "Noise augmentation level"
         },
         "steps": {
-          "node_id": 3,
+          "node_id": 5,
+          "widget_index": 2,
           "type": "integer",
-          "default": 30
+          "required": false,
+          "default": 30,
+          "min": 1,
+          "max": 150,
+          "description": "Sampling steps (30 recommended)"
         },
-        "frames": {
+        "cfg": {
-          "node_id": 3,
+          "node_id": 5,
-          "type": "integer",
+          "widget_index": 3,
-          "default": 25,
+          "type": "float",
-          "description": "Number of output frames"
+          "required": false,
+          "default": 6.0,
+          "min": 0.0,
+          "max": 30.0,
+          "description": "Classifier-free guidance scale"
         },
-        "motion_bucket": {
+        "seed": {
-          "node_id": 3,
+          "node_id": 5,
+          "widget_index": 0,
           "type": "integer",
-          "default": 127
+          "required": false,
+          "default": 42,
+          "description": "Random seed for reproducibility"
+        },
+        "output_fps": {
+          "node_id": 7,
+          "widget_index": 0,
+          "type": "integer",
+          "required": false,
+          "default": 6,
+          "description": "Output video framerate"
+        }
+      },
+      "outputs": {
+        "video": {
+          "node_id": 7,
+          "type": "video",
+          "format": "MP4 (H.264)",
+          "resolution": "1024x576 (configurable)",
+          "duration": "~4.2 seconds @ 6fps (25 frames)"
         }
       },
       "performance": {
         "avg_generation_time": "40-55 seconds",
-        "vram_usage": "~16-18GB",
+        "vram_usage": "~18-20GB",
-        "output": "25 frames (~4.2s @ 6fps)"
+        "gpu_utilization": "95-100%"
-      }
+      },
+      "use_cases": [
+        "Extended animations with smooth motion",
+        "Longer video loops from single frames",
+        "Cinematic camera movements",
+        "Product animation showcases"
+      ],
+      "notes": [
+        "SVD-XT generates 25 frames vs 14 frames in base SVD",
+        "Requires more VRAM (~18GB vs ~16GB)",
+        "Model auto-downloads on first use (~9GB)",
+        "Recommended resolution: 1024x576 (16:9)",
+        "Higher motion_bucket_id = more movement",
+        "Linear CFG guidance improves temporal consistency"
+      ]
     }
   },
   "version": 0.4
-}
+}

comfyui/workflows/nsfw/README.md

@@ -0,0 +1,373 @@
+# NSFW ComfyUI Workflows
+
+Production-ready workflows for NSFW content generation using CivitAI models.
+
+## Available Workflows
+
+### 1. LUSTIFY Realistic Text-to-Image
+**File:** `lustify-realistic-t2i-production-v1.json`
+
+**Model:** LUSTIFY v7.0 GGWP (`lustifySDXLNSFW_endgame.safetensors`)
+
+**Optimized Settings:**
+- Sampler: DPM++ 2M SDE
+- Scheduler: Exponential
+- Steps: 30
+- CFG Scale: 6.0
+- Resolution: 1024x1024 (SDXL)
+
+**Features:**
+- Photorealistic NSFW generation
+- Optimized negative prompts for LUSTIFY
+- Professional photography quality
+- Detailed skin texture and natural lighting
+- Supports both danbooru tags and natural language
+
+**Use Case:** High-quality photorealistic NSFW images focusing on women in various scenarios
+
+**Example Prompt:**
+```
+A beautiful woman in elegant dress, photorealistic, professional photography, high quality, detailed skin texture, natural lighting, 8k, masterpiece, best quality
+```
+
+**Negative Prompt:**
+```
+(worst quality, low quality:1.4), illustration, 3d, 2d, painting, cartoons, sketch, open mouth, bad anatomy, deformed, blurry, watermark, text
+```
+
+---
+
+### 2. Pony Diffusion Anime/Furry Text-to-Image
+**File:** `pony-anime-t2i-production-v1.json`
+
+**Model:** Pony Diffusion V6 XL (`add-detail-xl.safetensors`)
+
+**Optimized Settings:**
+- Sampler: Euler Ancestral
+- Scheduler: Normal
+- Steps: 35
+- CFG Scale: 7.5
+- Resolution: 1024x1024 (SDXL)
+
+**Features:**
+- Anime, cartoon, and furry NSFW generation
+- Danbooru tag support (optimized for tags)
+- Balanced content training (1:1:1 safe/questionable/explicit)
+- 2.6M aesthetically ranked training images
+- Versatile style support
+
+**Use Case:** High-quality anime/cartoon/furry NSFW content with tag-based prompting
+
+**Example Prompt (Danbooru Tags):**
+```
+1girl, solo, long_hair, breasts, smile, looking_at_viewer, anime style, high quality, detailed, masterpiece, best quality
+```
+
+**Example Prompt (Natural Language):**
+```
+Anime girl with long flowing hair, smiling and looking at viewer, detailed anime art style, high quality, masterpiece
+```
+
+**Negative Prompt:**
+```
+(worst quality, low quality:1.4), bad anatomy, deformed, blurry, watermark, text, signature, artist name
+```
+
+---
+
+### 3. RealVisXL Lightning Fast Text-to-Image
+**File:** `realvisxl-lightning-t2i-production-v1.json`
+
+**Model:** RealVisXL V5.0 Lightning (`realvisxlV50_v50LightningBakedvae.safetensors`)
+
+**Optimized Settings:**
+- Sampler: DPM++ SDE
+- Scheduler: Karras
+- Steps: 6 (Lightning fast!)
+- CFG Scale: 2.0
+- Resolution: 1024x1024 (SDXL)
+
+**Features:**
+- Ultra-fast 4-6 step generation (vs 30+ for standard models)
+- High-quality photorealistic output
+- Baked VAE for optimal quality
+- Professional portraits and scenes
+- Both SFW and NSFW capable
+
+**Use Case:** Rapid photorealistic NSFW generation when speed is priority
+
+**Example Prompt:**
+```
+Beautiful woman portrait, professional photography, natural lighting, high quality, detailed, 8k
+```
+
+**Negative Prompt:**
+```
+(worst quality, low quality:1.4), bad anatomy, deformed, blurry, watermark, text, ugly, cartoon, anime
+```
+
+---
+
+---
+
+## Advanced Workflows
+
+### 4. NSFW Ultimate SD Upscale
+**File:** `nsfw-ultimate-upscale-production-v1.json`
+
+**Model:** LUSTIFY v7.0 GGWP (`lustifySDXLNSFW_endgame.safetensors`)
+
+**Upscale Settings:**
+- Upscale Factor: 2x
+- Upscale Model: RealESRGAN_x2.pth
+- Denoise: 0.25 (preserves original composition)
+- Sampler: DPM++ 2M SDE
+- Scheduler: Exponential
+- Steps: 20
+- CFG Scale: 7.0
+- Tile Size: 512x512
+
+**Features:**
+- Professional 2x upscaling for NSFW images
+- Combines RealESRGAN with diffusion refinement
+- Tiled processing for handling large images
+- Optimized for detailed skin texture
+- Low denoise preserves original composition
+- Can swap checkpoint for different styles
+
+**Use Case:** Upscale NSFW images from 1024x1024 to 2048x2048 with enhanced detail
+
+**Enhancement Prompt:**
+```
+photorealistic, professional photography, high quality, detailed skin texture, natural lighting, 8k, masterpiece, best quality, sharp details
+```
+
+**Refinement Negative:**
+```
+(worst quality, low quality:1.4), blurry, pixelated, jpeg artifacts, bad anatomy, deformed, watermark, text
+```
+
+**Tips:**
+- Use denoise 0.2-0.3 for subtle enhancement
+- Match enhancement prompt to original generation prompt
+- Swap checkpoint node for different model styles (Pony, RealVisXL)
+- Processing time: ~90-150 seconds for 2x upscale
+
+---
+
+## Installation
+
+These workflows use models downloaded via `artifact_civitai_download.sh`:
+
+```bash
+# Models are already downloaded to /workspace/models/civitai/
+# and symlinked to /workspace/ComfyUI/models/checkpoints/
+
+# Verify models are available:
+ls -lh /workspace/ComfyUI/models/checkpoints/lustify*
+ls -lh /workspace/ComfyUI/models/checkpoints/*pony*
+ls -lh /workspace/ComfyUI/models/checkpoints/realvisx*
+```
+
+## Usage
+
+### Via ComfyUI API:
+
+```bash
+# Load workflow
+curl -X POST http://localhost:8188/api/prompt \
+  -H "Content-Type: application/json" \
+  -d @lustify-realistic-t2i-production-v1.json
+
+# Modify prompt and generate
+curl -X POST http://localhost:8188/api/prompt \
+  -H "Content-Type: application/json" \
+  -d '{
+    "prompt": {
+      "2": {
+        "inputs": {
+          "text": "Your custom NSFW prompt here"
+        }
+      }
+    }
+  }'
+```
+
+### Via ComfyUI UI:
+
+1. Open ComfyUI at `http://localhost:8188`
+2. Click "Load" and select workflow file
+3. Modify prompt in the "API Positive Prompt" node
+4. Click "Queue Prompt" to generate
+
+## Model Comparison
+
+| Model | Type | Speed | Quality | Best For | Prompt Style |
+|-------|------|-------|---------|----------|--------------|
+| LUSTIFY v7 | Photoreal | Medium (30 steps) | Excellent | Women, realistic scenes | Natural language |
+| Pony Diffusion v6 | Anime/Furry | Medium (35 steps) | Excellent | Anime, cartoon, furry | Danbooru tags |
+| RealVisXL Lightning | Photoreal | Very Fast (4-6 steps) | Excellent | Quick realistic generations | Natural language |
+
+## Tips
+
+### For LUSTIFY:
+- Use detailed prompts describing scene, lighting, quality
+- Avoid abstract/artistic terms
+- Include photography keywords: "professional photography", "8k", "detailed"
+- CFG 4-7 range works best
+
+### For Pony Diffusion:
+- Danbooru tags work very well (recommended over natural language)
+- Use tag format: `1girl, solo, breasts, smile, looking_at_viewer`
+- Mix safe/questionable/explicit content naturally
+- Can do anime, cartoon, and furry styles
+- Supports wide range of artistic styles
+- CFG 7-8.5 range works best
+
+### For RealVisXL Lightning:
+- Keep prompts concise for best results
+- Use lower CFG (1-2) for Lightning models
+- 4-6 steps is optimal balance
+- Great for batch generation
+
+## Advanced Usage
+
+### Adding LoRAs:
+Use the `../text-to-image/lora-fusion-t2i-production-v1.json` workflow for multi-LoRA stacking support.
+
+**How to use with NSFW models:**
+1. Load the LoRA fusion workflow
+2. Change checkpoint (node 1) to your NSFW model (LUSTIFY, Pony, RealVisXL)
+3. Stack up to 3 LoRAs with adjustable weights:
+   - LoRA 1 (Primary): 0.7-1.0 strength
+   - LoRA 2 (Secondary): 0.4-0.7 strength
+   - LoRA 3 (Accent): 0.2-0.5 strength
+4. Common use cases:
+   - Detail enhancer + skin texture + lighting style
+   - Character LoRA + pose enhancer + background style
+   - Primary art style + secondary blend + color grading
+
+### Upscaling:
+Use `nsfw-ultimate-upscale-production-v1.json` for professional 2x upscaling with LUSTIFY checkpoint.
+
+**Workflow:**
+1. Generate image with any NSFW T2I workflow (LUSTIFY, Pony, RealVisXL)
+2. Load the generated image into Ultimate SD Upscale workflow
+3. Use denoise 0.2-0.3 for subtle enhancement
+4. Match enhancement prompt to original generation prompt
+5. Output: 2048x2048 high-quality upscaled image
+
+### Style Transfer:
+Use IP-Adapter workflows in `../image-to-image/` directory to combine with style references.
+
+## Using Negative Embeddings
+
+Negative embeddings (textual inversions) are small trained models that significantly improve image quality when used in negative prompts. They help avoid common artifacts like bad anatomy, unrealistic features, and low-quality outputs.
+
+### Available Embeddings
+
+**For SDXL Models (LUSTIFY, RealVisXL):**
+- `BadX` - Fixes facial/hand artifacts, improves overall realism
+
+**For Pony/WAI-Illustrious Models:**
+- `zPDXL3` - Quality enhancement + removes censoring
+- `zPDXLxxx` - Enables explicit NSFW content
+
+**For SD 1.5 Models (future workflows):**
+- `BadDream` + `UnrealisticDream` - Dreamshaper artifacts + realism
+- `badhandv4` - Improves hand details
+- `FastNegativeV2` - General quality enhancement
+- `BadNegAnatomyV1-neg` - Better anatomy and body joints
+- `easynegative` - General-purpose quality
+
+### Installation
+
+Embeddings are automatically downloaded and linked to `/workspace/ComfyUI/models/embeddings/` when you run:
+
+```bash
+cd /workspace/ai && bash /workspace/bin/artifact_civitai_download.sh both -c models_civitai.yaml
+```
+
+### Usage Syntax
+
+**Basic usage in negative prompt:**
+```
+embedding:BadX
+embedding:zPDXL3
+```
+
+**Multiple embeddings:**
+```
+embedding:zPDXL3, embedding:zPDXLxxx, score_4, score_5, low quality
+```
+
+**With weight adjustment:**
+```
+(embedding:zPDXL3:1.5), (embedding:BadX.pt:1.2), ugly, deformed
+```
+
+**Note:** When using weights, include the full filename with extension (.pt or .safetensors).
+
+### Recommended Usage by Workflow
+
+**LUSTIFY Realistic T2I:**
+```
+Negative: embedding:BadX, (worst quality, low quality:1.4), bad anatomy, deformed, blurry, watermark, text
+```
+
+**RealVisXL Lightning T2I:**
+```
+Negative: embedding:BadX, (worst quality, low quality:1.4), ugly, cartoon, anime
+```
+
+**Pony Anime / WAI-Illustrious T2I:**
+```
+Negative: embedding:zPDXL3, embedding:zPDXLxxx, score_4, score_5, score_6, bad anatomy, deformed
+```
+
+Or with custom strength:
+```
+Negative: (embedding:zPDXL3:1.5), (embedding:zPDXLxxx:1.2), score_4, score_5, low quality
+```
+
+**NSFW Ultimate SD Upscale:**
+```
+Negative: embedding:BadX, (worst quality, low quality:1.4), blurry, pixelated, jpeg artifacts
+```
+
+### Important Compatibility Notes
+
+**CRITICAL:** SD 1.5 embeddings (BadDream, badhandv4, etc.) DO NOT work with SDXL models due to different CLIP architectures. Always use:
+- **SDXL models** → `BadX` embedding
+- **Pony models** → `zPDXL3` and `zPDXLxxx` embeddings
+- **SD 1.5 models** → SD 1.5 embeddings only
+
+Using the wrong embedding type will cause errors or unexpected results.
+
+### Tips
+
+- Start with recommended strength 1.0-2.0 for Pony embeddings
+- `zPDXL3` removes censoring, ideal for NSFW Pony workflows
+- Combine multiple embeddings for best results
+- `BadX` works with all general SDXL checkpoints (LUSTIFY, RealVisXL, etc.)
+- Higher CFG scales (≥11) work better with `badhandv4` (SD 1.5)
+
+## Notes
+
+- All workflows follow production naming standards
+- VRAM requirement: ~12GB per SDXL model
+- Can run one NSFW model at a time on 24GB GPU
+- Models automatically use BakedVAE when available
+- Face enhancement can be added via Impact-Pack nodes
+
+## Version History
+
+- v1.0 (2025-11-23): Initial release with complete NSFW workflow suite
+  - **Text-to-Image Workflows:**
+    - LUSTIFY realistic T2I (DPM++ 2M SDE, 30 steps, CFG 6.0)
+    - Pony Diffusion anime/furry T2I (Euler A, 35 steps, CFG 7.5)
+    - RealVisXL Lightning fast T2I (DPM++ SDE Karras, 6 steps, CFG 2.0)
+  - **Enhancement Workflows:**
+    - LoRA Fusion multi-stack (in ../text-to-image/, compatible with all NSFW checkpoints)
+    - NSFW Ultimate SD Upscale (2x upscaling with LUSTIFY + RealESRGAN)
+  - **Total:** 3 NSFW-specific workflows + 2 enhancement workflows = 5 production workflows

comfyui/workflows/nsfw/lustify-realistic-t2i-production-v1.json

@@ -0,0 +1,256 @@
+{
+  "last_node_id": 11,
+  "last_link_id": 16,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "CheckpointLoaderSimple",
+      "pos": [50, 100],
+      "size": {"0": 380, "1": 100},
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [2, 3],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [4],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": ["lustifySDXLNSFW_endgame.safetensors"],
+      "title": "LUSTIFY v7.0 GGWP Checkpoint"
+    },
+    {
+      "id": 2,
+      "type": "CLIPTextEncode",
+      "pos": [480, 100],
+      "size": {"0": 450, "1": 250},
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 2
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [5],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["A beautiful woman in elegant dress, photorealistic, professional photography, high quality, detailed skin texture, natural lighting, 8k, masterpiece, best quality"],
+      "title": "API Positive Prompt"
+    },
+    {
+      "id": 3,
+      "type": "CLIPTextEncode",
+      "pos": [480, 400],
+      "size": {"0": 450, "1": 150},
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["(worst quality, low quality:1.4), illustration, 3d, 2d, painting, cartoons, sketch, open mouth, bad anatomy, deformed, blurry, watermark, text"],
+      "title": "Negative Prompt (LUSTIFY Optimized)"
+    },
+    {
+      "id": 4,
+      "type": "EmptyLatentImage",
+      "pos": [480, 600],
+      "size": {"0": 315, "1": 110},
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [7],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "EmptyLatentImage"
+      },
+      "widgets_values": [1024, 1024, 1],
+      "title": "Latent Image (SDXL Resolution)"
+    },
+    {
+      "id": 5,
+      "type": "KSampler",
+      "pos": [980, 100],
+      "size": {"0": 350, "1": 290},
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 1
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 5
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 6
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [8],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        0,
+        "randomize",
+        30,
+        6.0,
+        "dpmpp_2m_sde",
+        "exponential",
+        1.0
+      ],
+      "title": "LUSTIFY Sampler (DPM++ 2M SDE / Exponential)"
+    },
+    {
+      "id": 6,
+      "type": "VAEDecode",
+      "pos": [1380, 100],
+      "size": {"0": 210, "1": 50},
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 8
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [9],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "title": "VAE Decode"
+    },
+    {
+      "id": 7,
+      "type": "SaveImage",
+      "pos": [1640, 100],
+      "size": {"0": 320, "1": 290},
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 9
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveImage"
+      },
+      "widgets_values": ["lustify_realistic"],
+      "title": "API Output Image"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 5, 0, "MODEL"],
+    [2, 1, 1, 2, 0, "CLIP"],
+    [3, 1, 1, 3, 0, "CLIP"],
+    [4, 1, 2, 6, 1, "VAE"],
+    [5, 2, 0, 5, 1, "CONDITIONING"],
+    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [7, 4, 0, 5, 3, "LATENT"],
+    [8, 5, 0, 6, 0, "LATENT"],
+    [9, 6, 0, 7, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "LUSTIFY Realistic NSFW Text-to-Image",
+      "version": "1.0",
+      "description": "Production workflow for LUSTIFY v7.0 GGWP checkpoint optimized for photorealistic NSFW generation with recommended settings (DPM++ 2M SDE, Exponential scheduler, 30 steps, CFG 6.0)",
+      "category": "nsfw",
+      "model": "lustifySDXLNSFW_endgame.safetensors",
+      "recommended_settings": {
+        "sampler": "dpmpp_2m_sde or dpmpp_3m_sde",
+        "scheduler": "exponential or karras",
+        "steps": 30,
+        "cfg_scale": "4.0-7.0"
+      },
+      "tags": ["nsfw", "realistic", "photography", "women", "sdxl"]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/nsfw/nsfw-ultimate-upscale-production-v1.json

@@ -0,0 +1,300 @@
+{
+  "last_node_id": 8,
+  "last_link_id": 10,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "LoadImage",
+      "pos": [50, 100],
+      "size": [350, 100],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null,
+          "slot_index": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoadImage"
+      },
+      "widgets_values": ["input_image.png", "image"],
+      "title": "API Input Image"
+    },
+    {
+      "id": 2,
+      "type": "CheckpointLoaderSimple",
+      "pos": [50, 250],
+      "size": [350, 100],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [2],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [3, 4],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [5],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": ["lustifySDXLNSFW_endgame.safetensors"],
+      "title": "LUSTIFY v7.0 GGWP Checkpoint"
+    },
+    {
+      "id": 3,
+      "type": "CLIPTextEncode",
+      "pos": [450, 100],
+      "size": [400, 200],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["photorealistic, professional photography, high quality, detailed skin texture, natural lighting, 8k, masterpiece, best quality, sharp details"],
+      "title": "API Positive Prompt (Enhancement)"
+    },
+    {
+      "id": 4,
+      "type": "CLIPTextEncode",
+      "pos": [450, 350],
+      "size": [400, 200],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [7],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["(worst quality, low quality:1.4), blurry, pixelated, jpeg artifacts, bad anatomy, deformed, watermark, text"],
+      "title": "API Negative Prompt (Refinement)"
+    },
+    {
+      "id": 8,
+      "type": "UpscaleModelLoader",
+      "pos": [50, 400],
+      "size": [350, 100],
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "UPSCALE_MODEL",
+          "type": "UPSCALE_MODEL",
+          "links": [10],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "UpscaleModelLoader"
+      },
+      "widgets_values": ["RealESRGAN_x2.pth"],
+      "title": "Upscale Model Loader (2x)"
+    },
+    {
+      "id": 5,
+      "type": "UltimateSDUpscale",
+      "pos": [900, 100],
+      "size": [315, 100],
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 1
+        },
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 2
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 6
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 7
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 5
+        },
+        {
+          "name": "upscale_model",
+          "type": "UPSCALE_MODEL",
+          "link": 10
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [8, 9],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "UltimateSDUpscale"
+      },
+      "widgets_values": [2, 42, "randomize", 20, 7.0, "dpmpp_2m_sde", "exponential", 0.25, "Linear", 512, 512, 8, 32, "None", 1.0, 64, 8, 16, true, false],
+      "title": "Ultimate SD Upscale NSFW (2x)"
+    },
+    {
+      "id": 6,
+      "type": "PreviewImage",
+      "pos": [1270, 100],
+      "size": [315, 100],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 8
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewImage"
+      },
+      "title": "Preview Output"
+    },
+    {
+      "id": 7,
+      "type": "SaveImage",
+      "pos": [1270, 400],
+      "size": [315, 100],
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 9
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveImage"
+      },
+      "widgets_values": ["nsfw_upscale_output"],
+      "title": "API Image Output"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 5, 0, "IMAGE"],
+    [2, 2, 0, 5, 1, "MODEL"],
+    [3, 2, 1, 3, 0, "CLIP"],
+    [4, 2, 1, 4, 0, "CLIP"],
+    [5, 2, 2, 5, 4, "VAE"],
+    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [7, 4, 0, 5, 3, "CONDITIONING"],
+    [8, 5, 0, 6, 0, "IMAGE"],
+    [9, 5, 0, 7, 0, "IMAGE"],
+    [10, 8, 0, 5, 5, "UPSCALE_MODEL"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "NSFW Ultimate SD Upscale Production",
+      "version": "1.0",
+      "description": "Professional 2x upscaling for NSFW images using Ultimate SD Upscale with LUSTIFY checkpoint. Combines RealESRGAN upscaling with diffusion refinement for superior detail, skin texture, and photorealistic quality in adult content.",
+      "category": "nsfw",
+      "model": "lustifySDXLNSFW_endgame.safetensors",
+      "recommended_settings": {
+        "upscale_factor": "2x (use RealESRGAN_x2.pth)",
+        "denoise": "0.2-0.3 (lower preserves original, higher adds detail)",
+        "sampler": "dpmpp_2m_sde",
+        "scheduler": "exponential",
+        "steps": 20,
+        "cfg_scale": "6.0-8.0",
+        "tile_size": "512x512 (optimal for SDXL)"
+      },
+      "features": [
+        "2x upscaling with RealESRGAN + diffusion refinement",
+        "LUSTIFY checkpoint for photorealistic NSFW enhancement",
+        "Tiled processing for handling large images",
+        "Optimized for detailed skin texture and natural lighting",
+        "Low denoise (0.25) preserves original composition",
+        "Can swap checkpoint for different styles (Pony, RealVisXL)"
+      ],
+      "tags": ["nsfw", "upscaling", "ultimate-sd-upscale", "lustify", "enhancement", "production"],
+      "requirements": {
+        "models": ["lustifySDXLNSFW_endgame.safetensors", "RealESRGAN_x2.pth"],
+        "custom_nodes": ["ComfyUI_UltimateSDUpscale"],
+        "vram_min": "18GB"
+      },
+      "usage_notes": {
+        "checkpoint_swap": "Change node 2 to use different NSFW models (pony, realvisxl, etc.)",
+        "denoise_tuning": "0.2 = subtle enhancement, 0.3 = moderate refinement, 0.4+ = creative upscaling",
+        "prompt_matching": "Match enhancement prompt to original generation prompt for best results",
+        "tile_overlap": "8px overlap prevents visible seams in large upscales"
+      },
+      "performance": {
+        "avg_generation_time": "90-150 seconds (1024x1024 -> 2048x2048)",
+        "vram_usage": "~20-22GB (LUSTIFY + tiled processing)"
+      }
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/nsfw/pony-anime-t2i-production-v1.json

@@ -0,0 +1,260 @@
+{
+  "last_node_id": 11,
+  "last_link_id": 16,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "CheckpointLoaderSimple",
+      "pos": [50, 100],
+      "size": {"0": 380, "1": 100},
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [2, 3],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [4],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": ["ponyDiffusionV6XL_v6StartWithThisOne.safetensors"],
+      "title": "Pony Diffusion V6 XL Checkpoint (Anime/Furry)"
+    },
+    {
+      "id": 2,
+      "type": "CLIPTextEncode",
+      "pos": [480, 100],
+      "size": {"0": 450, "1": 250},
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 2
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [5],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["1girl, solo, long_hair, breasts, smile, looking_at_viewer, anime style, high quality, detailed, masterpiece, best quality"],
+      "title": "API Positive Prompt (Danbooru Tags)"
+    },
+    {
+      "id": 3,
+      "type": "CLIPTextEncode",
+      "pos": [480, 400],
+      "size": {"0": 450, "1": 150},
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["(worst quality, low quality:1.4), bad anatomy, deformed, blurry, watermark, text, signature, artist name"],
+      "title": "Negative Prompt (Pony Optimized)"
+    },
+    {
+      "id": 4,
+      "type": "EmptyLatentImage",
+      "pos": [480, 600],
+      "size": {"0": 315, "1": 110},
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [7],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "EmptyLatentImage"
+      },
+      "widgets_values": [1024, 1024, 1],
+      "title": "Latent Image (SDXL Resolution)"
+    },
+    {
+      "id": 5,
+      "type": "KSampler",
+      "pos": [980, 100],
+      "size": {"0": 350, "1": 290},
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 1
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 5
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 6
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [8],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        0,
+        "randomize",
+        35,
+        7.5,
+        "euler_ancestral",
+        "normal",
+        1.0
+      ],
+      "title": "Pony Sampler (Euler A / Normal)"
+    },
+    {
+      "id": 6,
+      "type": "VAEDecode",
+      "pos": [1380, 100],
+      "size": {"0": 210, "1": 50},
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 8
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [9],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "title": "VAE Decode"
+    },
+    {
+      "id": 7,
+      "type": "SaveImage",
+      "pos": [1640, 100],
+      "size": {"0": 320, "1": 290},
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 9
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveImage"
+      },
+      "widgets_values": ["pony_anime"],
+      "title": "API Output Image"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 5, 0, "MODEL"],
+    [2, 1, 1, 2, 0, "CLIP"],
+    [3, 1, 1, 3, 0, "CLIP"],
+    [4, 1, 2, 6, 1, "VAE"],
+    [5, 2, 0, 5, 1, "CONDITIONING"],
+    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [7, 4, 0, 5, 3, "LATENT"],
+    [8, 5, 0, 6, 0, "LATENT"],
+    [9, 6, 0, 7, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "Pony Diffusion V6 XL Anime/Furry NSFW Text-to-Image",
+      "version": "1.0",
+      "description": "Production workflow for Pony Diffusion V6 XL optimized for anime, cartoon, and furry NSFW generation with danbooru tag support and balanced content (safe/questionable/explicit)",
+      "category": "nsfw",
+      "model": "ponyDiffusionV6XL_v6StartWithThisOne.safetensors",
+      "recommended_settings": {
+        "sampler": "euler_ancestral or dpmpp_2m",
+        "scheduler": "normal or karras",
+        "steps": "30-40",
+        "cfg_scale": "7.0-8.5"
+      },
+      "training_info": {
+        "images": "2.6M aesthetically ranked",
+        "ratio": "1:1:1 safe/questionable/explicit"
+      },
+      "tags": ["nsfw", "anime", "furry", "cartoon", "versatile", "danbooru", "sdxl"]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/nsfw/realvisxl-lightning-t2i-production-v1.json

@@ -0,0 +1,263 @@
+{
+  "last_node_id": 11,
+  "last_link_id": 16,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "CheckpointLoaderSimple",
+      "pos": [50, 100],
+      "size": {"0": 380, "1": 100},
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [2, 3],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [4],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": ["realvisxlV50_v50LightningBakedvae.safetensors"],
+      "title": "RealVisXL V5.0 Lightning Checkpoint"
+    },
+    {
+      "id": 2,
+      "type": "CLIPTextEncode",
+      "pos": [480, 100],
+      "size": {"0": 450, "1": 250},
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 2
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [5],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["Beautiful woman portrait, professional photography, natural lighting, high quality, detailed, 8k"],
+      "title": "API Positive Prompt"
+    },
+    {
+      "id": 3,
+      "type": "CLIPTextEncode",
+      "pos": [480, 400],
+      "size": {"0": 450, "1": 150},
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["(worst quality, low quality:1.4), bad anatomy, deformed, blurry, watermark, text, ugly, cartoon, anime"],
+      "title": "Negative Prompt (RealVisXL Optimized)"
+    },
+    {
+      "id": 4,
+      "type": "EmptyLatentImage",
+      "pos": [480, 600],
+      "size": {"0": 315, "1": 110},
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [7],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "EmptyLatentImage"
+      },
+      "widgets_values": [1024, 1024, 1],
+      "title": "Latent Image (SDXL Resolution)"
+    },
+    {
+      "id": 5,
+      "type": "KSampler",
+      "pos": [980, 100],
+      "size": {"0": 350, "1": 290},
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 1
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 5
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 6
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [8],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        0,
+        "randomize",
+        6,
+        2.0,
+        "dpmpp_sde",
+        "karras",
+        1.0
+      ],
+      "title": "Lightning Sampler (DPM++ SDE Karras / 6 Steps)"
+    },
+    {
+      "id": 6,
+      "type": "VAEDecode",
+      "pos": [1380, 100],
+      "size": {"0": 210, "1": 50},
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 8
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [9],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "title": "VAE Decode"
+    },
+    {
+      "id": 7,
+      "type": "SaveImage",
+      "pos": [1640, 100],
+      "size": {"0": 320, "1": 290},
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 9
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveImage"
+      },
+      "widgets_values": ["realvisxl_lightning"],
+      "title": "API Output Image"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 5, 0, "MODEL"],
+    [2, 1, 1, 2, 0, "CLIP"],
+    [3, 1, 1, 3, 0, "CLIP"],
+    [4, 1, 2, 6, 1, "VAE"],
+    [5, 2, 0, 5, 1, "CONDITIONING"],
+    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [7, 4, 0, 5, 3, "LATENT"],
+    [8, 5, 0, 6, 0, "LATENT"],
+    [9, 6, 0, 7, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "RealVisXL V5.0 Lightning Fast NSFW Text-to-Image",
+      "version": "1.0",
+      "description": "Production workflow for RealVisXL V5.0 Lightning checkpoint optimized for ultra-fast photorealistic NSFW generation with 4-6 step inference (vs 30+ for standard models)",
+      "category": "nsfw",
+      "model": "realvisxlV50_v50LightningBakedvae.safetensors",
+      "recommended_settings": {
+        "sampler": "dpmpp_sde with karras scheduler",
+        "scheduler": "karras",
+        "steps": "4-6 (Lightning optimized)",
+        "cfg_scale": "1.0-2.0 (lower for Lightning)"
+      },
+      "features": [
+        "Ultra-fast 4-6 step generation",
+        "High-quality photorealistic output",
+        "Baked VAE for optimal quality",
+        "Professional portraits and scenes",
+        "Both SFW and NSFW capable"
+      ],
+      "tags": ["nsfw", "realistic", "photorealistic", "lightning", "fast", "sdxl"]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-image/flux-dev-t2i-production-v1.json

@@ -1,12 +1,12 @@
 {
-  "last_node_id": 12,
+  "last_node_id": 9,
-  "last_link_id": 15,
+  "last_link_id": 13,
   "nodes": [
     {
       "id": 1,
-      "type": "CheckpointLoaderSimple",
+      "type": "UNETLoader",
       "pos": [50, 100],
-      "size": {"0": 350, "1": 100},
+      "size": [350, 100],
       "flags": {},
       "order": 0,
       "mode": 0,
@@ -16,34 +16,44 @@
           "type": "MODEL",
           "links": [1],
           "slot_index": 0
-        },
-        {
-          "name": "CLIP",
-          "type": "CLIP",
-          "links": [2, 3],
-          "slot_index": 1
-        },
-        {
-          "name": "VAE",
-          "type": "VAE",
-          "links": [4],
-          "slot_index": 2
         }
       ],
       "properties": {
-        "Node name for S&R": "CheckpointLoaderSimple"
+        "Node name for S&R": "UNETLoader"
       },
-      "widgets_values": ["diffusers/FLUX.1-dev"],
+      "widgets_values": ["flux1-dev.safetensors", "default"],
-      "title": "FLUX Dev Checkpoint Loader"
+      "title": "FLUX Dev UNET Loader"
     },
     {
       "id": 2,
-      "type": "CLIPTextEncode",
+      "type": "DualCLIPLoader",
-      "pos": [450, 100],
+      "pos": [50, 250],
-      "size": {"0": 400, "1": 200},
+      "size": [350, 100],
       "flags": {},
       "order": 1,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [2],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "DualCLIPLoader"
+      },
+      "widgets_values": ["t5xxl_fp16.safetensors", "clip_l.safetensors", "flux"],
+      "title": "FLUX CLIP Loader"
+    },
+    {
+      "id": 3,
+      "type": "CLIPTextEncodeFlux",
+      "pos": [450, 100],
+      "size": [400, 250],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
       "inputs": [
         {
           "name": "clip",
@@ -55,58 +65,61 @@
         {
           "name": "CONDITIONING",
           "type": "CONDITIONING",
-          "links": [5],
+          "links": [3, 8],
           "slot_index": 0
         }
       ],
       "properties": {
-        "Node name for S&R": "CLIPTextEncode"
+        "Node name for S&R": "CLIPTextEncodeFlux"
       },
-      "widgets_values": ["A beautiful mountain landscape at sunset, highly detailed, professional photography, cinematic lighting, 8k ultra HD"],
+      "widgets_values": [
+        "A beautiful mountain landscape at sunset, highly detailed, professional photography, cinematic lighting, 8k ultra HD",
+        "A beautiful mountain landscape at sunset, highly detailed, professional photography, cinematic lighting, 8k ultra HD",
+        3.5
+      ],
       "title": "API Positive Prompt"
     },
     {
-      "id": 3,
+      "id": 9,
-      "type": "CLIPTextEncode",
+      "type": "ConditioningZeroOut",
-      "pos": [450, 350],
+      "pos": [450, 400],
-      "size": {"0": 400, "1": 200},
+      "size": [315, 60],
       "flags": {},
-      "order": 2,
+      "order": 3,
       "mode": 0,
       "inputs": [
         {
-          "name": "clip",
+          "name": "conditioning",
-          "type": "CLIP",
+          "type": "CONDITIONING",
-          "link": 3
+          "link": 8
         }
       ],
       "outputs": [
         {
           "name": "CONDITIONING",
           "type": "CONDITIONING",
-          "links": [6],
+          "links": [9],
           "slot_index": 0
         }
       ],
       "properties": {
-        "Node name for S&R": "CLIPTextEncode"
+        "Node name for S&R": "ConditioningZeroOut"
       },
-      "widgets_values": ["blurry, low quality, distorted, watermark, text, signature, cartoon"],
+      "title": "Empty Negative Conditioning"
-      "title": "API Negative Prompt"
     },
     {
       "id": 4,
       "type": "EmptyLatentImage",
-      "pos": [450, 600],
+      "pos": [450, 500],
-      "size": {"0": 315, "1": 106},
+      "size": [315, 106],
       "flags": {},
-      "order": 3,
+      "order": 4,
       "mode": 0,
       "outputs": [
         {
           "name": "LATENT",
           "type": "LATENT",
-          "links": [7],
+          "links": [4],
           "slot_index": 0
         }
       ],
@@ -120,9 +133,9 @@
       "id": 5,
       "type": "KSampler",
       "pos": [900, 100],
-      "size": {"0": 315, "1": 474},
+      "size": [315, 474],
       "flags": {},
-      "order": 4,
+      "order": 5,
       "mode": 0,
       "inputs": [
         {
@@ -133,66 +146,80 @@
         {
           "name": "positive",
           "type": "CONDITIONING",
-          "link": 5
+          "link": 3
         },
         {
           "name": "negative",
           "type": "CONDITIONING",
-          "link": 6
+          "link": 9
         },
         {
           "name": "latent_image",
           "type": "LATENT",
-          "link": 7
+          "link": 4
         }
       ],
       "outputs": [
         {
           "name": "LATENT",
           "type": "LATENT",
-          "links": [8],
+          "links": [5],
           "slot_index": 0
         }
       ],
       "properties": {
         "Node name for S&R": "KSampler"
       },
-      "widgets_values": [
+      "widgets_values": [42, "fixed", 20, 1.0, "euler", "normal", 1.0],
-        42,
+      "title": "FLUX Dev Sampler (20 steps)"
-        "fixed",
-        20,
-        3.5,
-        "euler",
-        "normal",
-        1
-      ],
-      "title": "FLUX Dev Sampler (20-50 steps)"
     },
     {
       "id": 6,
+      "type": "VAELoader",
+      "pos": [900, 600],
+      "size": [315, 60],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAELoader"
+      },
+      "widgets_values": ["ae.safetensors"],
+      "title": "FLUX VAE Loader"
+    },
+    {
+      "id": 7,
       "type": "VAEDecode",
       "pos": [1270, 100],
-      "size": {"0": 210, "1": 46},
+      "size": [210, 46],
       "flags": {},
-      "order": 5,
+      "order": 7,
       "mode": 0,
       "inputs": [
         {
           "name": "samples",
           "type": "LATENT",
-          "link": 8
+          "link": 5
         },
         {
           "name": "vae",
           "type": "VAE",
-          "link": 4
+          "link": 6
         }
       ],
       "outputs": [
         {
           "name": "IMAGE",
           "type": "IMAGE",
-          "links": [9, 10],
+          "links": [7],
           "slot_index": 0
         }
       ],
@@ -201,39 +228,19 @@
       },
       "title": "VAE Decode"
     },
-    {
-      "id": 7,
-      "type": "PreviewImage",
-      "pos": [1530, 100],
-      "size": {"0": 400, "1": 400},
-      "flags": {},
-      "order": 6,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "images",
-          "type": "IMAGE",
-          "link": 9
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "PreviewImage"
-      },
-      "title": "Preview Output"
-    },
     {
       "id": 8,
       "type": "SaveImage",
-      "pos": [1530, 550],
+      "pos": [1530, 100],
-      "size": {"0": 400, "1": 100},
+      "size": [400, 100],
       "flags": {},
-      "order": 7,
+      "order": 8,
       "mode": 0,
       "inputs": [
         {
           "name": "images",
           "type": "IMAGE",
-          "link": 10
+          "link": 7
         }
       ],
       "properties": {
@@ -245,49 +252,39 @@
   ],
   "links": [
     [1, 1, 0, 5, 0, "MODEL"],
-    [2, 1, 1, 2, 0, "CLIP"],
+    [2, 2, 0, 3, 0, "CLIP"],
-    [3, 1, 1, 3, 0, "CLIP"],
+    [3, 3, 0, 5, 1, "CONDITIONING"],
-    [4, 1, 2, 6, 1, "VAE"],
+    [4, 4, 0, 5, 3, "LATENT"],
-    [5, 2, 0, 5, 1, "CONDITIONING"],
+    [5, 5, 0, 7, 0, "LATENT"],
-    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [6, 6, 0, 7, 1, "VAE"],
-    [7, 4, 0, 5, 3, "LATENT"],
+    [7, 7, 0, 8, 0, "IMAGE"],
-    [8, 5, 0, 6, 0, "LATENT"],
+    [8, 3, 0, 9, 0, "CONDITIONING"],
-    [9, 6, 0, 7, 0, "IMAGE"],
+    [9, 9, 0, 5, 2, "CONDITIONING"]
-    [10, 6, 0, 8, 0, "IMAGE"]
   ],
   "groups": [],
   "config": {},
   "extra": {
     "workflow_info": {
       "name": "FLUX Dev Text-to-Image Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
-      "author": "RunPod AI Model Orchestrator",
       "description": "High-quality text-to-image generation using FLUX.1-dev (20-50 steps). Optimized for final production outputs with excellent detail and coherence.",
       "category": "text-to-image",
       "tags": ["flux", "dev", "high-quality", "production", "t2i"],
       "requirements": {
-        "models": ["FLUX.1-dev"],
+        "models": ["FLUX.1-dev", "CLIP-L", "T5-XXL FP16"],
         "custom_nodes": [],
         "vram_min": "20GB",
         "vram_recommended": "24GB"
       },
       "parameters": {
         "prompt": {
-          "node_id": 2,
+          "node_id": 3,
           "widget_index": 0,
           "type": "string",
           "required": true,
           "default": "A beautiful mountain landscape at sunset",
           "description": "Text description of desired image"
         },
-        "negative_prompt": {
-          "node_id": 3,
-          "widget_index": 0,
-          "type": "string",
-          "required": false,
-          "default": "blurry, low quality",
-          "description": "Undesired elements to avoid"
-        },
         "width": {
           "node_id": 4,
           "widget_index": 0,
@@ -328,15 +325,15 @@
           "max": 50,
           "description": "Number of sampling steps (20-50 recommended for FLUX Dev)"
         },
-        "cfg": {
+        "guidance": {
-          "node_id": 5,
+          "node_id": 3,
-          "widget_index": 3,
+          "widget_index": 2,
           "type": "float",
           "required": false,
           "default": 3.5,
           "min": 1.0,
           "max": 10.0,
-          "description": "Classifier-free guidance scale (3.5 recommended)"
+          "description": "Guidance scale (3.5 recommended for FLUX)"
         }
       },
       "outputs": {
@@ -351,7 +348,14 @@
         "avg_generation_time": "25-35 seconds (20 steps), 60-75 seconds (50 steps)",
         "vram_usage": "~20-22GB",
         "gpu_utilization": "95-100%"
-      }
+      },
+      "notes": [
+        "FLUX uses UNETLoader instead of CheckpointLoader",
+        "DualCLIPLoader loads both T5-XXL and CLIP-L text encoders",
+        "CLIPTextEncodeFlux is FLUX-specific text encoder",
+        "No negative prompt for FLUX - guidance is handled differently",
+        "CFG scale of 1.0 recommended, guidance in text encoder"
+      ]
     }
   },
   "version": 0.4

comfyui/workflows/text-to-image/flux-schnell-t2i-production-v1.json

@@ -1,12 +1,12 @@
 {
-  "last_node_id": 12,
+  "last_node_id": 9,
-  "last_link_id": 15,
+  "last_link_id": 9,
   "nodes": [
     {
       "id": 1,
-      "type": "CheckpointLoaderSimple",
+      "type": "UNETLoader",
       "pos": [50, 100],
-      "size": {"0": 350, "1": 100},
+      "size": [350, 100],
       "flags": {},
       "order": 0,
       "mode": 0,
@@ -16,34 +16,44 @@
           "type": "MODEL",
           "links": [1],
           "slot_index": 0
-        },
-        {
-          "name": "CLIP",
-          "type": "CLIP",
-          "links": [2, 3],
-          "slot_index": 1
-        },
-        {
-          "name": "VAE",
-          "type": "VAE",
-          "links": [4],
-          "slot_index": 2
         }
       ],
       "properties": {
-        "Node name for S&R": "CheckpointLoaderSimple"
+        "Node name for S&R": "UNETLoader"
       },
-      "widgets_values": ["diffusers/FLUX.1-schnell"],
+      "widgets_values": ["flux1-schnell.safetensors", "default"],
-      "title": "FLUX Schnell Checkpoint Loader"
+      "title": "FLUX Schnell UNET Loader"
     },
     {
       "id": 2,
-      "type": "CLIPTextEncode",
+      "type": "DualCLIPLoader",
-      "pos": [450, 100],
+      "pos": [50, 250],
-      "size": {"0": 400, "1": 200},
+      "size": [350, 100],
       "flags": {},
       "order": 1,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [2],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "DualCLIPLoader"
+      },
+      "widgets_values": ["t5xxl_fp16.safetensors", "clip_l.safetensors", "flux"],
+      "title": "FLUX CLIP Loader"
+    },
+    {
+      "id": 3,
+      "type": "CLIPTextEncodeFlux",
+      "pos": [450, 100],
+      "size": [400, 250],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
       "inputs": [
         {
           "name": "clip",
@@ -55,58 +65,61 @@
         {
           "name": "CONDITIONING",
           "type": "CONDITIONING",
-          "links": [5],
+          "links": [3, 8],
           "slot_index": 0
         }
       ],
       "properties": {
-        "Node name for S&R": "CLIPTextEncode"
+        "Node name for S&R": "CLIPTextEncodeFlux"
       },
-      "widgets_values": ["A beautiful mountain landscape at sunset, highly detailed, cinematic lighting, 8k"],
+      "widgets_values": [
+        "A beautiful mountain landscape at sunset, highly detailed, cinematic lighting, 8k",
+        "A beautiful mountain landscape at sunset, highly detailed, cinematic lighting, 8k",
+        3.5
+      ],
       "title": "API Positive Prompt"
     },
     {
-      "id": 3,
+      "id": 9,
-      "type": "CLIPTextEncode",
+      "type": "ConditioningZeroOut",
-      "pos": [450, 350],
+      "pos": [450, 400],
-      "size": {"0": 400, "1": 200},
+      "size": [315, 60],
       "flags": {},
-      "order": 2,
+      "order": 3,
       "mode": 0,
       "inputs": [
         {
-          "name": "clip",
+          "name": "conditioning",
-          "type": "CLIP",
+          "type": "CONDITIONING",
-          "link": 3
+          "link": 8
         }
       ],
       "outputs": [
         {
           "name": "CONDITIONING",
           "type": "CONDITIONING",
-          "links": [6],
+          "links": [9],
           "slot_index": 0
         }
       ],
       "properties": {
-        "Node name for S&R": "CLIPTextEncode"
+        "Node name for S&R": "ConditioningZeroOut"
       },
-      "widgets_values": ["blurry, low quality, distorted, watermark"],
+      "title": "Empty Negative Conditioning"
-      "title": "API Negative Prompt"
     },
     {
       "id": 4,
       "type": "EmptyLatentImage",
-      "pos": [450, 600],
+      "pos": [450, 500],
-      "size": {"0": 315, "1": 106},
+      "size": [315, 106],
       "flags": {},
-      "order": 3,
+      "order": 4,
       "mode": 0,
       "outputs": [
         {
           "name": "LATENT",
           "type": "LATENT",
-          "links": [7],
+          "links": [4],
           "slot_index": 0
         }
       ],
@@ -120,9 +133,9 @@
       "id": 5,
       "type": "KSampler",
       "pos": [900, 100],
-      "size": {"0": 315, "1": 474},
+      "size": [315, 474],
       "flags": {},
-      "order": 4,
+      "order": 5,
       "mode": 0,
       "inputs": [
         {
@@ -133,66 +146,80 @@
         {
           "name": "positive",
           "type": "CONDITIONING",
-          "link": 5
+          "link": 3
         },
         {
           "name": "negative",
           "type": "CONDITIONING",
-          "link": 6
+          "link": 9
         },
         {
           "name": "latent_image",
           "type": "LATENT",
-          "link": 7
+          "link": 4
         }
       ],
       "outputs": [
         {
           "name": "LATENT",
           "type": "LATENT",
-          "links": [8],
+          "links": [5],
           "slot_index": 0
         }
       ],
       "properties": {
         "Node name for S&R": "KSampler"
       },
-      "widgets_values": [
+      "widgets_values": [42, "fixed", 4, 1.0, "euler", "normal", 1.0],
-        42,
-        "fixed",
-        4,
-        1.0,
-        "euler",
-        "normal",
-        1
-      ],
       "title": "FLUX Schnell Sampler (4 steps)"
     },
     {
       "id": 6,
+      "type": "VAELoader",
+      "pos": [900, 600],
+      "size": [315, 60],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAELoader"
+      },
+      "widgets_values": ["ae.safetensors"],
+      "title": "FLUX VAE Loader"
+    },
+    {
+      "id": 7,
       "type": "VAEDecode",
       "pos": [1270, 100],
-      "size": {"0": 210, "1": 46},
+      "size": [210, 46],
       "flags": {},
-      "order": 5,
+      "order": 7,
       "mode": 0,
       "inputs": [
         {
           "name": "samples",
           "type": "LATENT",
-          "link": 8
+          "link": 5
         },
         {
           "name": "vae",
           "type": "VAE",
-          "link": 4
+          "link": 6
         }
       ],
       "outputs": [
         {
           "name": "IMAGE",
           "type": "IMAGE",
-          "links": [9, 10],
+          "links": [7],
           "slot_index": 0
         }
       ],
@@ -201,39 +228,19 @@
       },
       "title": "VAE Decode"
     },
-    {
-      "id": 7,
-      "type": "PreviewImage",
-      "pos": [1530, 100],
-      "size": {"0": 400, "1": 400},
-      "flags": {},
-      "order": 6,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "images",
-          "type": "IMAGE",
-          "link": 9
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "PreviewImage"
-      },
-      "title": "Preview Output"
-    },
     {
       "id": 8,
       "type": "SaveImage",
-      "pos": [1530, 550],
+      "pos": [1530, 100],
-      "size": {"0": 400, "1": 100},
+      "size": [400, 100],
       "flags": {},
-      "order": 7,
+      "order": 8,
       "mode": 0,
       "inputs": [
         {
           "name": "images",
           "type": "IMAGE",
-          "link": 10
+          "link": 7
         }
       ],
       "properties": {
@@ -245,49 +252,39 @@
   ],
   "links": [
     [1, 1, 0, 5, 0, "MODEL"],
-    [2, 1, 1, 2, 0, "CLIP"],
+    [2, 2, 0, 3, 0, "CLIP"],
-    [3, 1, 1, 3, 0, "CLIP"],
+    [3, 3, 0, 5, 1, "CONDITIONING"],
-    [4, 1, 2, 6, 1, "VAE"],
+    [4, 4, 0, 5, 3, "LATENT"],
-    [5, 2, 0, 5, 1, "CONDITIONING"],
+    [5, 5, 0, 7, 0, "LATENT"],
-    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [6, 6, 0, 7, 1, "VAE"],
-    [7, 4, 0, 5, 3, "LATENT"],
+    [7, 7, 0, 8, 0, "IMAGE"],
-    [8, 5, 0, 6, 0, "LATENT"],
+    [8, 3, 0, 9, 0, "CONDITIONING"],
-    [9, 6, 0, 7, 0, "IMAGE"],
+    [9, 9, 0, 5, 2, "CONDITIONING"]
-    [10, 6, 0, 8, 0, "IMAGE"]
   ],
   "groups": [],
   "config": {},
   "extra": {
     "workflow_info": {
       "name": "FLUX Schnell Text-to-Image Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
-      "author": "RunPod AI Model Orchestrator",
       "description": "Fast text-to-image generation using FLUX.1-schnell (4 steps). Optimized for rapid prototyping and iteration.",
       "category": "text-to-image",
       "tags": ["flux", "schnell", "fast", "production", "t2i"],
       "requirements": {
-        "models": ["FLUX.1-schnell"],
+        "models": ["FLUX.1-schnell", "CLIP-L", "T5-XXL FP16"],
         "custom_nodes": [],
         "vram_min": "16GB",
         "vram_recommended": "24GB"
       },
       "parameters": {
         "prompt": {
-          "node_id": 2,
+          "node_id": 3,
           "widget_index": 0,
           "type": "string",
           "required": true,
           "default": "A beautiful mountain landscape at sunset",
           "description": "Text description of desired image"
         },
-        "negative_prompt": {
-          "node_id": 3,
-          "widget_index": 0,
-          "type": "string",
-          "required": false,
-          "default": "blurry, low quality",
-          "description": "Undesired elements to avoid"
-        },
         "width": {
           "node_id": 4,
           "widget_index": 0,
@@ -326,17 +323,17 @@
           "default": 4,
           "min": 1,
           "max": 10,
-          "description": "Number of sampling steps (4 recommended for Schnell)"
+          "description": "Number of sampling steps (4 recommended for FLUX Schnell)"
         },
-        "cfg": {
+        "guidance": {
-          "node_id": 5,
+          "node_id": 3,
-          "widget_index": 3,
+          "widget_index": 2,
           "type": "float",
           "required": false,
-          "default": 1.0,
+          "default": 3.5,
           "min": 1.0,
-          "max": 1.0,
+          "max": 10.0,
-          "description": "Classifier-free guidance scale (1.0 for FLUX Schnell)"
+          "description": "Guidance scale (3.5 recommended for FLUX)"
         }
       },
       "outputs": {
@@ -348,10 +345,18 @@
         }
       },
       "performance": {
-        "avg_generation_time": "5-8 seconds",
+        "avg_generation_time": "5-8 seconds (4 steps)",
-        "vram_usage": "~16GB",
+        "vram_usage": "~16-18GB",
         "gpu_utilization": "95-100%"
-      }
+      },
+      "notes": [
+        "FLUX uses UNETLoader instead of CheckpointLoader",
+        "DualCLIPLoader loads both T5-XXL and CLIP-L text encoders",
+        "CLIPTextEncodeFlux is FLUX-specific text encoder",
+        "No negative prompt for FLUX - guidance is handled differently",
+        "CFG scale of 1.0 recommended, guidance in text encoder",
+        "Schnell is optimized for 4 steps - fastest FLUX variant"
+      ]
     }
   },
   "version": 0.4

comfyui/workflows/text-to-image/lora-fusion-t2i-production-v1.json

@@ -0,0 +1,408 @@
+{
+  "last_node_id": 15,
+  "last_link_id": 24,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "CheckpointLoaderSimple",
+      "pos": [50, 100],
+      "size": {"0": 380, "1": 100},
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [2],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [3],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": ["sd_xl_base_1.0.safetensors"],
+      "title": "Base Checkpoint Loader"
+    },
+    {
+      "id": 2,
+      "type": "LoraLoader",
+      "pos": [480, 100],
+      "size": {"0": 380, "1": 130},
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 1
+        },
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 2
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [4],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [5],
+          "slot_index": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoraLoader"
+      },
+      "widgets_values": ["add-detail-xl.safetensors", 0.8, 0.8],
+      "title": "LoRA 1 (Primary Style)"
+    },
+    {
+      "id": 3,
+      "type": "LoraLoader",
+      "pos": [920, 100],
+      "size": {"0": 380, "1": 130},
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 4
+        },
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [6],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [7],
+          "slot_index": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoraLoader"
+      },
+      "widgets_values": ["add-detail-xl.safetensors", 0.5, 0.5],
+      "title": "LoRA 2 (Secondary Style)"
+    },
+    {
+      "id": 4,
+      "type": "LoraLoader",
+      "pos": [1360, 100],
+      "size": {"0": 380, "1": 130},
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 6
+        },
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [8],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [9, 10],
+          "slot_index": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoraLoader"
+      },
+      "widgets_values": ["add-detail-xl.safetensors", 0.3, 0.3],
+      "title": "LoRA 3 (Accent/Detail)"
+    },
+    {
+      "id": 5,
+      "type": "CLIPTextEncode",
+      "pos": [50, 280],
+      "size": {"0": 450, "1": 250},
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 9
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [11],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["A beautiful scenic landscape with mountains, vibrant colors, detailed, high quality, masterpiece, best quality, 8k"],
+      "title": "API Positive Prompt"
+    },
+    {
+      "id": 6,
+      "type": "CLIPTextEncode",
+      "pos": [50, 580],
+      "size": {"0": 450, "1": 150},
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 10
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [12],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": ["(worst quality, low quality:1.4), bad anatomy, deformed, blurry, watermark, text"],
+      "title": "Negative Prompt"
+    },
+    {
+      "id": 7,
+      "type": "EmptyLatentImage",
+      "pos": [560, 580],
+      "size": {"0": 315, "1": 110},
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [13],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "EmptyLatentImage"
+      },
+      "widgets_values": [1024, 1024, 1],
+      "title": "Latent Image (SDXL Resolution)"
+    },
+    {
+      "id": 8,
+      "type": "KSampler",
+      "pos": [920, 280],
+      "size": {"0": 350, "1": 290},
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 8
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 11
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 12
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 13
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [14],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        0,
+        "randomize",
+        30,
+        7.0,
+        "dpmpp_2m_sde",
+        "karras",
+        1.0
+      ],
+      "title": "KSampler (Standard Settings)"
+    },
+    {
+      "id": 9,
+      "type": "VAEDecode",
+      "pos": [1320, 280],
+      "size": {"0": 210, "1": 50},
+      "flags": {},
+      "order": 8,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 14
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [15],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "title": "VAE Decode"
+    },
+    {
+      "id": 10,
+      "type": "SaveImage",
+      "pos": [1580, 280],
+      "size": {"0": 320, "1": 290},
+      "flags": {},
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 15
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveImage"
+      },
+      "widgets_values": ["lora_fusion"],
+      "title": "API Output Image"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 2, 0, "MODEL"],
+    [2, 1, 1, 2, 1, "CLIP"],
+    [3, 1, 2, 9, 1, "VAE"],
+    [4, 2, 0, 3, 0, "MODEL"],
+    [5, 2, 1, 3, 1, "CLIP"],
+    [6, 3, 0, 4, 0, "MODEL"],
+    [7, 3, 1, 4, 1, "CLIP"],
+    [8, 4, 0, 8, 0, "MODEL"],
+    [9, 4, 1, 5, 0, "CLIP"],
+    [10, 4, 1, 6, 0, "CLIP"],
+    [11, 5, 0, 8, 1, "CONDITIONING"],
+    [12, 6, 0, 8, 2, "CONDITIONING"],
+    [13, 7, 0, 8, 3, "LATENT"],
+    [14, 8, 0, 9, 0, "LATENT"],
+    [15, 9, 0, 10, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "LoRA Fusion Multi-Stack Text-to-Image",
+      "version": "1.0",
+      "description": "Production workflow for combining multiple LoRA models with adjustable weights. Stack up to 3 LoRAs with fine-tuned strength control for style mixing and enhancement. Works with any SDXL checkpoint including NSFW models.",
+      "category": "text-to-image",
+      "features": [
+        "Triple LoRA stacking with independent weight control",
+        "Compatible with any SDXL checkpoint (SFW/NSFW)",
+        "Hierarchical strength: Primary (0.7-1.0), Secondary (0.4-0.7), Accent (0.2-0.5)",
+        "Supports style LoRAs, detail enhancers, and character LoRAs",
+        "Easy to disable individual LoRAs by setting weights to 0.0"
+      ],
+      "recommended_usage": {
+        "lora_1_strength": "0.7-1.0 (primary style influence)",
+        "lora_2_strength": "0.4-0.7 (secondary style blend)",
+        "lora_3_strength": "0.2-0.5 (subtle accent/details)",
+        "disable_lora": "Set both model and clip weights to 0.0"
+      },
+      "examples": [
+        {
+          "use_case": "Portrait Enhancement",
+          "lora_1": "Detail enhancer (0.8)",
+          "lora_2": "Skin texture LoRA (0.5)",
+          "lora_3": "Lighting style (0.3)"
+        },
+        {
+          "use_case": "Artistic Style Fusion",
+          "lora_1": "Primary art style (1.0)",
+          "lora_2": "Secondary style blend (0.6)",
+          "lora_3": "Color grading (0.4)"
+        },
+        {
+          "use_case": "Character + Style",
+          "lora_1": "Character LoRA (0.9)",
+          "lora_2": "Pose/anatomy enhancer (0.5)",
+          "lora_3": "Background style (0.3)"
+        }
+      ],
+      "tags": ["lora", "multi-stack", "fusion", "style-mixing", "enhancement", "sdxl"]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-image/sd35-large-t2i-production-v1.json

@@ -1,12 +1,12 @@
 {
-  "last_node_id": 12,
+  "last_node_id": 9,
-  "last_link_id": 15,
+  "last_link_id": 10,
   "nodes": [
     {
       "id": 1,
       "type": "CheckpointLoaderSimple",
       "pos": [50, 100],
-      "size": {"0": 350, "1": 100},
+      "size": [350, 100],
       "flags": {},
       "order": 0,
       "mode": 0,
@@ -20,8 +20,7 @@
         {
           "name": "CLIP",
           "type": "CLIP",
-          "links": [2, 3],
+          "links": null
-          "slot_index": 1
         },
         {
           "name": "VAE",
@@ -33,22 +32,44 @@
       "properties": {
         "Node name for S&R": "CheckpointLoaderSimple"
       },
-      "widgets_values": ["diffusers/stable-diffusion-3.5-large"],
+      "widgets_values": ["sd3.5_large.safetensors"],
       "title": "SD3.5 Large Checkpoint Loader"
     },
+    {
+      "id": 9,
+      "type": "TripleCLIPLoader",
+      "pos": [50, 250],
+      "size": [350, 150],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [9, 10],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "TripleCLIPLoader"
+      },
+      "widgets_values": ["clip_l.safetensors", "clip_g.safetensors", "t5xxl_fp16.safetensors"],
+      "title": "SD3.5 CLIP Loaders (L+G+T5)"
+    },
     {
       "id": 2,
       "type": "CLIPTextEncode",
       "pos": [450, 100],
-      "size": {"0": 400, "1": 200},
+      "size": [400, 200],
       "flags": {},
-      "order": 1,
+      "order": 2,
       "mode": 0,
       "inputs": [
         {
           "name": "clip",
           "type": "CLIP",
-          "link": 2
+          "link": 9
         }
       ],
       "outputs": [
@@ -69,15 +90,15 @@
       "id": 3,
       "type": "CLIPTextEncode",
       "pos": [450, 350],
-      "size": {"0": 400, "1": 200},
+      "size": [400, 200],
       "flags": {},
-      "order": 2,
+      "order": 3,
       "mode": 0,
       "inputs": [
         {
           "name": "clip",
           "type": "CLIP",
-          "link": 3
+          "link": 10
         }
       ],
       "outputs": [
@@ -98,9 +119,9 @@
       "id": 4,
       "type": "EmptyLatentImage",
       "pos": [450, 600],
-      "size": {"0": 315, "1": 106},
+      "size": [315, 106],
       "flags": {},
-      "order": 3,
+      "order": 4,
       "mode": 0,
       "outputs": [
         {
@@ -120,9 +141,9 @@
       "id": 5,
       "type": "KSampler",
       "pos": [900, 100],
-      "size": {"0": 315, "1": 474},
+      "size": [315, 474],
       "flags": {},
-      "order": 4,
+      "order": 5,
       "mode": 0,
       "inputs": [
         {
@@ -157,24 +178,16 @@
       "properties": {
         "Node name for S&R": "KSampler"
       },
-      "widgets_values": [
+      "widgets_values": [42, "fixed", 28, 4.5, "dpmpp_2m", "karras", 1],
-        42,
-        "fixed",
-        28,
-        4.5,
-        "dpmpp_2m",
-        "karras",
-        1
-      ],
       "title": "SD3.5 Sampler (28 steps)"
     },
     {
       "id": 6,
       "type": "VAEDecode",
       "pos": [1270, 100],
-      "size": {"0": 210, "1": 46},
+      "size": [210, 46],
       "flags": {},
-      "order": 5,
+      "order": 6,
       "mode": 0,
       "inputs": [
         {
@@ -192,7 +205,7 @@
         {
           "name": "IMAGE",
           "type": "IMAGE",
-          "links": [9, 10],
+          "links": [2, 3],
           "slot_index": 0
         }
       ],
@@ -205,15 +218,15 @@
       "id": 7,
       "type": "PreviewImage",
       "pos": [1530, 100],
-      "size": {"0": 400, "1": 400},
+      "size": [400, 400],
       "flags": {},
-      "order": 6,
+      "order": 7,
       "mode": 0,
       "inputs": [
         {
           "name": "images",
           "type": "IMAGE",
-          "link": 9
+          "link": 2
         }
       ],
       "properties": {
@@ -225,15 +238,15 @@
       "id": 8,
       "type": "SaveImage",
       "pos": [1530, 550],
-      "size": {"0": 400, "1": 100},
+      "size": [400, 100],
       "flags": {},
-      "order": 7,
+      "order": 8,
       "mode": 0,
       "inputs": [
         {
           "name": "images",
           "type": "IMAGE",
-          "link": 10
+          "link": 3
         }
       ],
       "properties": {
@@ -245,28 +258,27 @@
   ],
   "links": [
     [1, 1, 0, 5, 0, "MODEL"],
-    [2, 1, 1, 2, 0, "CLIP"],
+    [2, 6, 0, 7, 0, "IMAGE"],
-    [3, 1, 1, 3, 0, "CLIP"],
+    [3, 6, 0, 8, 0, "IMAGE"],
     [4, 1, 2, 6, 1, "VAE"],
     [5, 2, 0, 5, 1, "CONDITIONING"],
     [6, 3, 0, 5, 2, "CONDITIONING"],
     [7, 4, 0, 5, 3, "LATENT"],
     [8, 5, 0, 6, 0, "LATENT"],
-    [9, 6, 0, 7, 0, "IMAGE"],
+    [9, 9, 0, 2, 0, "CLIP"],
-    [10, 6, 0, 8, 0, "IMAGE"]
+    [10, 9, 0, 3, 0, "CLIP"]
   ],
   "groups": [],
   "config": {},
   "extra": {
     "workflow_info": {
       "name": "Stable Diffusion 3.5 Large Text-to-Image Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
-      "author": "RunPod AI Model Orchestrator",
       "description": "Latest generation text-to-image using Stable Diffusion 3.5 Large (28 steps). Provides excellent photorealism and prompt adherence.",
       "category": "text-to-image",
       "tags": ["sd3.5", "stable-diffusion", "large", "production", "t2i", "photorealistic"],
       "requirements": {
-        "models": ["stable-diffusion-3.5-large"],
+        "models": ["SD3.5-large", "CLIP-L", "CLIP-G", "T5-XXL FP16"],
         "custom_nodes": [],
         "vram_min": "18GB",
         "vram_recommended": "24GB"
@@ -351,7 +363,13 @@
         "avg_generation_time": "35-45 seconds",
         "vram_usage": "~18-20GB",
         "gpu_utilization": "95-100%"
-      }
+      },
+      "notes": [
+        "SD3.5 uses CheckpointLoaderSimple for MODEL+VAE",
+        "TripleCLIPLoader loads CLIP-L, CLIP-G, and T5-XXL separately",
+        "Checkpoint file doesn't include CLIP encoders",
+        "All three CLIP encoders required for best quality"
+      ]
     }
   },
   "version": 0.4

comfyui/workflows/text-to-image/sdxl-refiner-t2i-production-v1.json

@@ -33,7 +33,7 @@
       "properties": {
         "Node name for S&R": "CheckpointLoaderSimple"
       },
-      "widgets_values": ["diffusers/stable-diffusion-xl-base-1.0"],
+      "widgets_values": ["sd_xl_base_1.0.safetensors"],
       "title": "SDXL Base Checkpoint Loader"
     },
     {
@@ -67,7 +67,7 @@
       "properties": {
         "Node name for S&R": "CheckpointLoaderSimple"
       },
-      "widgets_values": ["diffusers/stable-diffusion-xl-refiner-1.0"],
+      "widgets_values": ["sd_xl_refiner_1.0.safetensors"],
       "title": "SDXL Refiner Checkpoint Loader"
     },
     {

comfyui/workflows/text-to-music/acestep-m2m-editing-v1.json

@@ -0,0 +1,865 @@
+{
+  "id": "88ac5dad-efd7-40bb-84fe-fbaefdee1fa9",
+  "revision": 0,
+  "last_node_id": 75,
+  "last_link_id": 138,
+  "nodes": [
+    {
+      "id": 49,
+      "type": "LatentApplyOperationCFG",
+      "pos": [
+        940,
+        -160
+      ],
+      "size": [
+        290,
+        50
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 10,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 113
+        },
+        {
+          "name": "operation",
+          "type": "LATENT_OPERATION",
+          "link": 114
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            121
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "LatentApplyOperationCFG"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 40,
+      "type": "CheckpointLoaderSimple",
+      "pos": [
+        180,
+        -160
+      ],
+      "size": [
+        370,
+        98
+      ],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            115
+          ]
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [
+            80
+          ]
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [
+            83,
+            137
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "CheckpointLoaderSimple",
+        "models": [
+          {
+            "name": "ace_step_v1_3.5b.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/resolve/main/all_in_one/ace_step_v1_3.5b.safetensors?download=true",
+            "directory": "checkpoints"
+          }
+        ]
+      },
+      "widgets_values": [
+        "ace_step_v1_3.5b.safetensors"
+      ],
+      "color": "#322",
+      "bgcolor": "#533"
+    },
+    {
+      "id": 48,
+      "type": "MarkdownNote",
+      "pos": [
+        -460,
+        -200
+      ],
+      "size": [
+        610,
+        820
+      ],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "title": "About ACE Step and Multi-language Input",
+      "properties": {},
+      "widgets_values": [
+        "[Tutorial](http://docs.comfy.org/tutorials/audio/ace-step/ace-step-v1) | [教程](http://docs.comfy.org/zh-CN/tutorials/audio/ace-step/ace-step-v1)\n\n\n### Model Download\n\nDownload the following model and save it to the **ComfyUI/models/checkpoints** folder.\n[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors)\n\n\n### Multilingual Support\n\nCurrently, the implementation of multi-language support for ACE-Step V1 is achieved by uniformly converting different languages into English characters. At present, in ComfyUI, we haven't implemented the step of converting multi-languages into English. This is because if we need to implement the corresponding conversion, we have to add additional core dependencies of ComfyUI, which may lead to uncertain dependency conflicts.\n\nSo, currently, if you need to input multi-language text, you have to manually convert it into English characters to complete this process. Then, at the beginning of the corresponding `lyrics`, input the abbreviation of the corresponding language code.\n\nFor example, for Chinese, use `[zh]`, for Japanese use `[ja]`, for Korean use `[ko]`, and so on. For specific language input, please check the examples in the instructions. \n\nFor example, Chinese `[zh]`, Japanese `[ja]`, Korean `[ko]`, etc.\n\nExample:\n\n```\n[verse]\n\n[zh]wo3zou3guo4shen1ye4de5jie1dao4\n[zh]leng3feng1chui1luan4si1nian4de5piao4liang4wai4tao4\n[zh]ni3de5wei1xiao4xiang4xing1guang1hen3xuan4yao4\n[zh]zhao4liang4le5wo3gu1du2de5mei3fen1mei3miao3\n\n[chorus]\n\n[verse]​\n[ko]hamkke si-kkeuleo-un sesang-ui sodong-eul pihae​\n[ko]honja ogsang-eseo dalbich-ui eolyeompus-ileul balaboda​\n[ko]niga salang-eun lideum-i ganghan eum-ag gatdago malhaess-eo​\n[ko]han ta han tamada ma-eum-ui ondoga eolmana heojeonhanji ijge hae\n\n[bridge]\n[es]cantar mi anhelo por ti sin ocultar\n[es]como poesía y pintura, lleno de anhelo indescifrable\n[es]tu sombra es tan terca como el viento, inborrable\n[es]persiguiéndote en vuelo, brilla como cruzar una mar de nubes\n\n[chorus]\n[fr]que tu sois le vent qui souffle sur ma main\n[fr]un contact chaud comme la douce pluie printanière\n[fr]que tu sois le vent qui s'entoure de mon corps\n[fr]un amour profond qui ne s'éloignera jamais\n\n```\n\n---\n\n### 模型下载\n\n下载下面的模型并保存到 **ComfyUI/models/checkpoints** 文件夹下\n[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors)\n\n\n### 多语言支持\n\n目前 ACE-Step V1 多语言的实现是通过将不同语言统一转换为英文字符来实现的，目前在 ComfyUI 中我们并没有实现多语言转换为英文的这一步骤。因为如果需要实现对应转换，则需要增加额外的 ComfyUI 核心依赖，这将可能带来不确定的依赖冲突。\n\n所以目前如果你需要输入多语言，则需要手动转换为英文字符来实现这一过程，然后在对应 `lyrics` 开头输入对应语言代码的缩写。\n\n比如中文`[zh]` 日语 `[ja]` 韩语 `[ko]` 等，具体语言输入请查看说明中的示例\n\n"
+      ],
+      "color": "#432",
+      "bgcolor": "#653"
+    },
+    {
+      "id": 18,
+      "type": "VAEDecodeAudio",
+      "pos": [
+        1080,
+        270
+      ],
+      "size": [
+        150.93612670898438,
+        46
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 13,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 122
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 83
+        }
+      ],
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [
+            126,
+            127,
+            128
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "VAEDecodeAudio"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 60,
+      "type": "SaveAudio",
+      "pos": [
+        1260,
+        40
+      ],
+      "size": [
+        610,
+        112
+      ],
+      "flags": {},
+      "order": 15,
+      "mode": 4,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 127
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "SaveAudio"
+      },
+      "widgets_values": [
+        "audio/ComfyUI"
+      ]
+    },
+    {
+      "id": 61,
+      "type": "SaveAudioOpus",
+      "pos": [
+        1260,
+        220
+      ],
+      "size": [
+        610,
+        136
+      ],
+      "flags": {},
+      "order": 16,
+      "mode": 4,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 128
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "SaveAudioOpus"
+      },
+      "widgets_values": [
+        "audio/ComfyUI",
+        "128k"
+      ]
+    },
+    {
+      "id": 44,
+      "type": "ConditioningZeroOut",
+      "pos": [
+        600,
+        70
+      ],
+      "size": [
+        197.712890625,
+        26
+      ],
+      "flags": {
+        "collapsed": true
+      },
+      "order": 11,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "conditioning",
+          "type": "CONDITIONING",
+          "link": 108
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            120
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "ConditioningZeroOut"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 51,
+      "type": "ModelSamplingSD3",
+      "pos": [
+        590,
+        -40
+      ],
+      "size": [
+        330,
+        60
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 115
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            113
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "ModelSamplingSD3"
+      },
+      "widgets_values": [
+        5.000000000000001
+      ]
+    },
+    {
+      "id": 50,
+      "type": "LatentOperationTonemapReinhard",
+      "pos": [
+        590,
+        -160
+      ],
+      "size": [
+        330,
+        58
+      ],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "LATENT_OPERATION",
+          "type": "LATENT_OPERATION",
+          "links": [
+            114
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "LatentOperationTonemapReinhard"
+      },
+      "widgets_values": [
+        1.0000000000000002
+      ]
+    },
+    {
+      "id": 17,
+      "type": "EmptyAceStepLatentAudio",
+      "pos": [
+        180,
+        50
+      ],
+      "size": [
+        370,
+        82
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 4,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": []
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "EmptyAceStepLatentAudio"
+      },
+      "widgets_values": [
+        120,
+        1
+      ]
+    },
+    {
+      "id": 68,
+      "type": "VAEEncodeAudio",
+      "pos": [
+        180,
+        180
+      ],
+      "size": [
+        370,
+        46
+      ],
+      "flags": {},
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 136
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 137
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            138
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "VAEEncodeAudio"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 64,
+      "type": "LoadAudio",
+      "pos": [
+        180,
+        340
+      ],
+      "size": [
+        370,
+        140
+      ],
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [
+            136
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "LoadAudio"
+      },
+      "widgets_values": [
+        "audio_ace_step_1_t2a_song-1.mp3",
+        null,
+        null
+      ],
+      "color": "#322",
+      "bgcolor": "#533"
+    },
+    {
+      "id": 52,
+      "type": "KSampler",
+      "pos": [
+        940,
+        -40
+      ],
+      "size": [
+        290,
+        262
+      ],
+      "flags": {},
+      "order": 12,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 121
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 117
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 120
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 138
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "slot_index": 0,
+          "links": [
+            122
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        938549746349002,
+        "randomize",
+        50,
+        5,
+        "euler",
+        "simple",
+        0.30000000000000004
+      ]
+    },
+    {
+      "id": 59,
+      "type": "SaveAudioMP3",
+      "pos": [
+        1260,
+        -160
+      ],
+      "size": [
+        610,
+        136
+      ],
+      "flags": {},
+      "order": 14,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 126
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "SaveAudioMP3"
+      },
+      "widgets_values": [
+        "audio/ComfyUI",
+        "V0"
+      ]
+    },
+    {
+      "id": 73,
+      "type": "Note",
+      "pos": [
+        1260,
+        410
+      ],
+      "size": [
+        610,
+        90
+      ],
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "properties": {},
+      "widgets_values": [
+        "These nodes can save audio in different formats. Currently, all the modes are Bypass. You can enable them as per your needs.\n\n这些节点可以将 audio  保存成不同格式，目前的模式都是 Bypass ，你可以按你的需要来启用"
+      ],
+      "color": "#432",
+      "bgcolor": "#653"
+    },
+    {
+      "id": 14,
+      "type": "TextEncodeAceStepAudio",
+      "pos": [
+        590,
+        120
+      ],
+      "size": [
+        340,
+        500
+      ],
+      "flags": {},
+      "order": 8,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 80
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            108,
+            117
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "TextEncodeAceStepAudio"
+      },
+      "widgets_values": [
+        "anime, cute female vocals, kawaii pop, j-pop, childish, piano, guitar, synthesizer, fast, happy, cheerful, lighthearted",
+        "[verse]\nフワフワ　オミミガ\nユレルヨ　カゼノナカ\nキラキラ　アオイメ\nミツメル　セカイヲ\n\n[verse]\nフワフワ　シッポハ\nオオキク　ユレルヨ\nキンイロ　カミノケ\nナビクヨ　カゼノナカ\n\n[verse]\nコンフィーユーアイノ\nマモリビト\nピンクノ　セーターデ\nエガオヲ　クレルヨ\n\nアオイロ　スカートト\nクロイコート　キンノモヨウ\nヤサシイ　ヒカリガ\nツツムヨ　フェネックガール\n\n[verse]\nフワフワ　オミミデ\nキコエル　ココロノ　コエ\nダイスキ　フェネックガール\nイツデモ　ソバニイルヨ",
+        0.9900000000000002
+      ]
+    },
+    {
+      "id": 75,
+      "type": "MarkdownNote",
+      "pos": [
+        950,
+        410
+      ],
+      "size": [
+        280,
+        210
+      ],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "title": "About Repainting",
+      "properties": {},
+      "widgets_values": [
+        "Providing the lyrics of the original song or the modified lyrics is very important for the output of repainting or editing. \n\nAdjust the value of the **denoise** parameter in KSampler. The larger the value, the lower the similarity between the output audio and the original audio.\n\n提供原始歌曲的歌词或者修改后的歌词对于音频编辑的输出是非常重要的，调整 KSampler 中的  denoise 参数的数值，数值越大输出的音频与原始音频相似度越低"
+      ],
+      "color": "#432",
+      "bgcolor": "#653"
+    }
+  ],
+  "links": [
+    [
+      80,
+      40,
+      1,
+      14,
+      0,
+      "CLIP"
+    ],
+    [
+      83,
+      40,
+      2,
+      18,
+      1,
+      "VAE"
+    ],
+    [
+      108,
+      14,
+      0,
+      44,
+      0,
+      "CONDITIONING"
+    ],
+    [
+      113,
+      51,
+      0,
+      49,
+      0,
+      "MODEL"
+    ],
+    [
+      114,
+      50,
+      0,
+      49,
+      1,
+      "LATENT_OPERATION"
+    ],
+    [
+      115,
+      40,
+      0,
+      51,
+      0,
+      "MODEL"
+    ],
+    [
+      117,
+      14,
+      0,
+      52,
+      1,
+      "CONDITIONING"
+    ],
+    [
+      120,
+      44,
+      0,
+      52,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      121,
+      49,
+      0,
+      52,
+      0,
+      "MODEL"
+    ],
+    [
+      122,
+      52,
+      0,
+      18,
+      0,
+      "LATENT"
+    ],
+    [
+      126,
+      18,
+      0,
+      59,
+      0,
+      "AUDIO"
+    ],
+    [
+      127,
+      18,
+      0,
+      60,
+      0,
+      "AUDIO"
+    ],
+    [
+      128,
+      18,
+      0,
+      61,
+      0,
+      "AUDIO"
+    ],
+    [
+      136,
+      64,
+      0,
+      68,
+      0,
+      "AUDIO"
+    ],
+    [
+      137,
+      40,
+      2,
+      68,
+      1,
+      "VAE"
+    ],
+    [
+      138,
+      68,
+      0,
+      52,
+      3,
+      "LATENT"
+    ]
+  ],
+  "groups": [
+    {
+      "id": 1,
+      "title": "Load model here",
+      "bounding": [
+        170,
+        -230,
+        390,
+        180
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 4,
+      "title": "Latent",
+      "bounding": [
+        170,
+        -30,
+        390,
+        280
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 5,
+      "title": "Adjust the vocal volume",
+      "bounding": [
+        580,
+        -230,
+        350,
+        140
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 6,
+      "title": "For repainting",
+      "bounding": [
+        170,
+        270,
+        390,
+        223.60000610351562
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 7,
+      "title": "Output",
+      "bounding": [
+        1250,
+        -230,
+        630,
+        760
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    }
+  ],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 0.6830134553650705,
+      "offset": [
+        785.724285521853,
+        434.02395631202546
+      ]
+    },
+    "frontendVersion": "1.19.9",
+    "node_versions": {
+      "comfy-core": "0.3.34",
+      "ace-step": "06f751d65491c9077fa2bc9b06d2c6f2a90e4c56"
+    },
+    "VHS_latentpreview": false,
+    "VHS_latentpreviewrate": 0,
+    "VHS_MetadataImage": true,
+    "VHS_KeepIntermediate": true
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-music/acestep-official-t2m-v1.json

@@ -0,0 +1,841 @@
+{
+  "id": "88ac5dad-efd7-40bb-84fe-fbaefdee1fa9",
+  "revision": 0,
+  "last_node_id": 73,
+  "last_link_id": 137,
+  "nodes": [
+    {
+      "id": 49,
+      "type": "LatentApplyOperationCFG",
+      "pos": [
+        940,
+        -160
+      ],
+      "size": [
+        290,
+        50
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 113
+        },
+        {
+          "name": "operation",
+          "type": "LATENT_OPERATION",
+          "link": 114
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            121
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "LatentApplyOperationCFG"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 64,
+      "type": "LoadAudio",
+      "pos": [
+        180,
+        340
+      ],
+      "size": [
+        370,
+        140
+      ],
+      "flags": {},
+      "order": 0,
+      "mode": 4,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [
+            136
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "LoadAudio"
+      },
+      "widgets_values": [
+        "ace_step_example.flac",
+        null,
+        null
+      ],
+      "color": "#322",
+      "bgcolor": "#533"
+    },
+    {
+      "id": 68,
+      "type": "VAEEncodeAudio",
+      "pos": [
+        180,
+        180
+      ],
+      "size": [
+        370,
+        46
+      ],
+      "flags": {},
+      "order": 8,
+      "mode": 4,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 136
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 137
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": null
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "VAEEncodeAudio"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 40,
+      "type": "CheckpointLoaderSimple",
+      "pos": [
+        180,
+        -160
+      ],
+      "size": [
+        370,
+        98
+      ],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            115
+          ]
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [
+            80
+          ]
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [
+            83,
+            137
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "CheckpointLoaderSimple",
+        "models": [
+          {
+            "name": "ace_step_v1_3.5b.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/resolve/main/all_in_one/ace_step_v1_3.5b.safetensors?download=true",
+            "directory": "checkpoints"
+          }
+        ]
+      },
+      "widgets_values": [
+        "ace_step_v1_3.5b.safetensors"
+      ],
+      "color": "#322",
+      "bgcolor": "#533"
+    },
+    {
+      "id": 48,
+      "type": "MarkdownNote",
+      "pos": [
+        -460,
+        -200
+      ],
+      "size": [
+        610,
+        820
+      ],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "title": "About ACE Step and Multi-language Input",
+      "properties": {},
+      "widgets_values": [
+        "[Tutorial](http://docs.comfy.org/tutorials/audio/ace-step/ace-step-v1) | [教程](http://docs.comfy.org/zh-CN/tutorials/audio/ace-step/ace-step-v1)\n\n\n### Model Download\n\nDownload the following model and save it to the **ComfyUI/models/checkpoints** folder.\n[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors)\n\n\n### Multilingual Support\n\nCurrently, the implementation of multi-language support for ACE-Step V1 is achieved by uniformly converting different languages into English characters. At present, in ComfyUI, we haven't implemented the step of converting multi-languages into English. This is because if we need to implement the corresponding conversion, we have to add additional core dependencies of ComfyUI, which may lead to uncertain dependency conflicts.\n\nSo, currently, if you need to input multi-language text, you have to manually convert it into English characters to complete this process. Then, at the beginning of the corresponding `lyrics`, input the abbreviation of the corresponding language code.\n\nFor example, for Chinese, use `[zh]`, for Japanese use `[ja]`, for Korean use `[ko]`, and so on. For specific language input, please check the examples in the instructions. \n\nFor example, Chinese `[zh]`, Japanese `[ja]`, Korean `[ko]`, etc.\n\nExample:\n\n```\n[verse]\n\n[zh]wo3zou3guo4shen1ye4de5jie1dao4\n[zh]leng3feng1chui1luan4si1nian4de5piao4liang4wai4tao4\n[zh]ni3de5wei1xiao4xiang4xing1guang1hen3xuan4yao4\n[zh]zhao4liang4le5wo3gu1du2de5mei3fen1mei3miao3\n\n[chorus]\n\n[verse]​\n[ko]hamkke si-kkeuleo-un sesang-ui sodong-eul pihae​\n[ko]honja ogsang-eseo dalbich-ui eolyeompus-ileul balaboda​\n[ko]niga salang-eun lideum-i ganghan eum-ag gatdago malhaess-eo​\n[ko]han ta han tamada ma-eum-ui ondoga eolmana heojeonhanji ijge hae\n\n[bridge]\n[es]cantar mi anhelo por ti sin ocultar\n[es]como poesía y pintura, lleno de anhelo indescifrable\n[es]tu sombra es tan terca como el viento, inborrable\n[es]persiguiéndote en vuelo, brilla como cruzar una mar de nubes\n\n[chorus]\n[fr]que tu sois le vent qui souffle sur ma main\n[fr]un contact chaud comme la douce pluie printanière\n[fr]que tu sois le vent qui s'entoure de mon corps\n[fr]un amour profond qui ne s'éloignera jamais\n\n```\n\n---\n\n### 模型下载\n\n下载下面的模型并保存到 **ComfyUI/models/checkpoints** 文件夹下\n[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors)\n\n\n### 多语言支持\n\n目前 ACE-Step V1 多语言的实现是通过将不同语言统一转换为英文字符来实现的，目前在 ComfyUI 中我们并没有实现多语言转换为英文的这一步骤。因为如果需要实现对应转换，则需要增加额外的 ComfyUI 核心依赖，这将可能带来不确定的依赖冲突。\n\n所以目前如果你需要输入多语言，则需要手动转换为英文字符来实现这一过程，然后在对应 `lyrics` 开头输入对应语言代码的缩写。\n\n比如中文`[zh]` 日语 `[ja]` 韩语 `[ko]` 等，具体语言输入请查看说明中的示例\n\n"
+      ],
+      "color": "#432",
+      "bgcolor": "#653"
+    },
+    {
+      "id": 18,
+      "type": "VAEDecodeAudio",
+      "pos": [
+        1080,
+        270
+      ],
+      "size": [
+        150.93612670898438,
+        46
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 12,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 122
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 83
+        }
+      ],
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [
+            126,
+            127,
+            128
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "VAEDecodeAudio"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 60,
+      "type": "SaveAudio",
+      "pos": [
+        1260,
+        40
+      ],
+      "size": [
+        610,
+        112
+      ],
+      "flags": {},
+      "order": 14,
+      "mode": 4,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 127
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "SaveAudio"
+      },
+      "widgets_values": [
+        "audio/ComfyUI"
+      ]
+    },
+    {
+      "id": 61,
+      "type": "SaveAudioOpus",
+      "pos": [
+        1260,
+        220
+      ],
+      "size": [
+        610,
+        136
+      ],
+      "flags": {},
+      "order": 15,
+      "mode": 4,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 128
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "SaveAudioOpus"
+      },
+      "widgets_values": [
+        "audio/ComfyUI",
+        "128k"
+      ]
+    },
+    {
+      "id": 73,
+      "type": "Note",
+      "pos": [
+        1260,
+        410
+      ],
+      "size": [
+        610,
+        90
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "properties": {},
+      "widgets_values": [
+        "These nodes can save audio in different formats. Currently, all the modes are Bypass. You can enable them as per your needs.\n\n这些节点可以将 audio  保存成不同格式，目前的模式都是 Bypass ，你可以按你的需要来启用"
+      ],
+      "color": "#432",
+      "bgcolor": "#653"
+    },
+    {
+      "id": 44,
+      "type": "ConditioningZeroOut",
+      "pos": [
+        600,
+        70
+      ],
+      "size": [
+        197.712890625,
+        26
+      ],
+      "flags": {
+        "collapsed": true
+      },
+      "order": 10,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "conditioning",
+          "type": "CONDITIONING",
+          "link": 108
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            120
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "ConditioningZeroOut"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 51,
+      "type": "ModelSamplingSD3",
+      "pos": [
+        590,
+        -40
+      ],
+      "size": [
+        330,
+        60
+      ],
+      "flags": {
+        "collapsed": false
+      },
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 115
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            113
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "ModelSamplingSD3"
+      },
+      "widgets_values": [
+        5.000000000000001
+      ]
+    },
+    {
+      "id": 50,
+      "type": "LatentOperationTonemapReinhard",
+      "pos": [
+        590,
+        -160
+      ],
+      "size": [
+        330,
+        58
+      ],
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "LATENT_OPERATION",
+          "type": "LATENT_OPERATION",
+          "links": [
+            114
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "LatentOperationTonemapReinhard"
+      },
+      "widgets_values": [
+        1.0000000000000002
+      ]
+    },
+    {
+      "id": 52,
+      "type": "KSampler",
+      "pos": [
+        940,
+        -40
+      ],
+      "size": [
+        290,
+        262
+      ],
+      "flags": {},
+      "order": 11,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 121
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 117
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 120
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 119
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "slot_index": 0,
+          "links": [
+            122
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "KSampler"
+      },
+      "widgets_values": [
+        468254064217846,
+        "randomize",
+        50,
+        5,
+        "euler",
+        "simple",
+        1
+      ]
+    },
+    {
+      "id": 14,
+      "type": "TextEncodeAceStepAudio",
+      "pos": [
+        590,
+        120
+      ],
+      "size": [
+        340,
+        500
+      ],
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 80
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            108,
+            117
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "TextEncodeAceStepAudio"
+      },
+      "widgets_values": [
+        "anime, soft female vocals, kawaii pop, j-pop, childish, piano, guitar, synthesizer, fast, happy, cheerful, lighthearted\t\n",
+        "[inst]\n\n[verse]\nふわふわ　おみみが\nゆれるよ　かぜのなか\nきらきら　あおいめ\nみつめる　せかいを\n\n[verse]\nふわふわ　しっぽは\nおおきく　ゆれるよ\nきんいろ　かみのけ\nなびくよ　かぜのなか\n\n[verse]\nコンフィーユーアイの\nまもりびと\nピンクの　セーターで\nえがおを　くれるよ\n\nあおいろ　スカートと\nくろいコート　きんのもよう\nやさしい　ひかりが\nつつむよ　フェネックガール\n\n[verse]\nふわふわ　おみみで\nきこえる　こころの　こえ\nだいすき　フェネックガール\nいつでも　そばにいるよ\n\n\n",
+        0.9900000000000002
+      ]
+    },
+    {
+      "id": 17,
+      "type": "EmptyAceStepLatentAudio",
+      "pos": [
+        180,
+        50
+      ],
+      "size": [
+        370,
+        82
+      ],
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            119
+          ]
+        }
+      ],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.32",
+        "Node name for S&R": "EmptyAceStepLatentAudio"
+      },
+      "widgets_values": [
+        120,
+        1
+      ]
+    },
+    {
+      "id": 59,
+      "type": "SaveAudioMP3",
+      "pos": [
+        1260,
+        -160
+      ],
+      "size": [
+        610,
+        136
+      ],
+      "flags": {},
+      "order": 13,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 126
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "cnr_id": "comfy-core",
+        "ver": "0.3.34",
+        "Node name for S&R": "SaveAudioMP3"
+      },
+      "widgets_values": [
+        "audio/ComfyUI",
+        "V0"
+      ]
+    }
+  ],
+  "links": [
+    [
+      80,
+      40,
+      1,
+      14,
+      0,
+      "CLIP"
+    ],
+    [
+      83,
+      40,
+      2,
+      18,
+      1,
+      "VAE"
+    ],
+    [
+      108,
+      14,
+      0,
+      44,
+      0,
+      "CONDITIONING"
+    ],
+    [
+      113,
+      51,
+      0,
+      49,
+      0,
+      "MODEL"
+    ],
+    [
+      114,
+      50,
+      0,
+      49,
+      1,
+      "LATENT_OPERATION"
+    ],
+    [
+      115,
+      40,
+      0,
+      51,
+      0,
+      "MODEL"
+    ],
+    [
+      117,
+      14,
+      0,
+      52,
+      1,
+      "CONDITIONING"
+    ],
+    [
+      119,
+      17,
+      0,
+      52,
+      3,
+      "LATENT"
+    ],
+    [
+      120,
+      44,
+      0,
+      52,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      121,
+      49,
+      0,
+      52,
+      0,
+      "MODEL"
+    ],
+    [
+      122,
+      52,
+      0,
+      18,
+      0,
+      "LATENT"
+    ],
+    [
+      126,
+      18,
+      0,
+      59,
+      0,
+      "AUDIO"
+    ],
+    [
+      127,
+      18,
+      0,
+      60,
+      0,
+      "AUDIO"
+    ],
+    [
+      128,
+      18,
+      0,
+      61,
+      0,
+      "AUDIO"
+    ],
+    [
+      136,
+      64,
+      0,
+      68,
+      0,
+      "AUDIO"
+    ],
+    [
+      137,
+      40,
+      2,
+      68,
+      1,
+      "VAE"
+    ]
+  ],
+  "groups": [
+    {
+      "id": 1,
+      "title": "Load model here",
+      "bounding": [
+        170,
+        -230,
+        390,
+        180
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 4,
+      "title": "Latent",
+      "bounding": [
+        170,
+        -30,
+        390,
+        280
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 5,
+      "title": "Adjust the vocal volume",
+      "bounding": [
+        580,
+        -230,
+        350,
+        140
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 6,
+      "title": "For repainting",
+      "bounding": [
+        170,
+        270,
+        390,
+        223.60000610351562
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    },
+    {
+      "id": 7,
+      "title": "Output",
+      "bounding": [
+        1250,
+        -230,
+        630,
+        760
+      ],
+      "color": "#3f789e",
+      "font_size": 24,
+      "flags": {}
+    }
+  ],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 1,
+      "offset": [
+        -147.02717343600432,
+        384.62272311479
+      ]
+    },
+    "frontendVersion": "1.19.9",
+    "node_versions": {
+      "comfy-core": "0.3.34",
+      "ace-step": "06f751d65491c9077fa2bc9b06d2c6f2a90e4c56"
+    },
+    "VHS_latentpreview": false,
+    "VHS_latentpreviewrate": 0,
+    "VHS_MetadataImage": true,
+    "VHS_KeepIntermediate": true
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-music/diffrhythm-full-length-t2m-v1.json

@@ -0,0 +1,130 @@
+{
+  "last_node_id": 3,
+  "last_link_id": 2,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "DiffRhythmRun",
+      "pos": [100, 100],
+      "size": [400, 400],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [1, 2]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "DiffRhythmRun"
+      },
+      "widgets_values": [
+        "cfm_full_model.pt",
+        "Cinematic orchestral piece with soaring strings, powerful brass, and emotional piano melodies building to an epic crescendo",
+        true,
+        "euler",
+        30,
+        4,
+        "quality",
+        123,
+        "randomize",
+        false,
+        "[-1, 20], [60, -1]"
+      ],
+      "title": "DiffRhythm Full-Length Text-to-Music (4m45s)"
+    },
+    {
+      "id": 2,
+      "type": "PreviewAudio",
+      "pos": [600, 100],
+      "size": [300, 100],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewAudio"
+      },
+      "title": "Preview Audio"
+    },
+    {
+      "id": 3,
+      "type": "SaveAudio",
+      "pos": [600, 250],
+      "size": [300, 100],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveAudio"
+      },
+      "widgets_values": [
+        "diffrhythm_full_output"
+      ],
+      "title": "Save Audio"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"],
+    [2, 1, 0, 3, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "DiffRhythm Full-Length Text-to-Music v1",
+      "description": "Full-length music generation using DiffRhythm Full (4 minutes 45 seconds)",
+      "version": "1.0.0",
+      "author": "valknar@pivoine.art",
+      "category": "text-to-music",
+      "tags": ["diffrhythm", "music-generation", "text-to-music", "full-length", "4m45s"],
+      "requirements": {
+        "custom_nodes": ["ComfyUI_DiffRhythm"],
+        "models": ["ASLP-lab/DiffRhythm-full", "ASLP-lab/DiffRhythm-vae", "OpenMuQ/MuQ-MuLan-large", "OpenMuQ/MuQ-large-msd-iter", "FacebookAI/xlm-roberta-base"],
+        "vram_min": "16GB",
+        "vram_recommended": "20GB",
+        "system_deps": ["espeak-ng"]
+      },
+      "usage": {
+        "model": "cfm_full_model.pt (DiffRhythm Full - 4m45s/285s generation)",
+        "style_prompt": "Detailed text description of the desired full-length music composition",
+        "unload_model": "Boolean to unload model after generation (default: true)",
+        "odeint_method": "ODE solver: euler, midpoint, rk4, implicit_adams (default: euler)",
+        "steps": "Number of diffusion steps: 1-100 (default: 30)",
+        "cfg": "Classifier-free guidance scale: 1-10 (default: 4)",
+        "quality_or_speed": "Generation mode: quality or speed (default: quality for full-length)",
+        "seed": "Random seed for reproducibility (default: 123)",
+        "edit": "Enable segment editing mode (default: false)",
+        "edit_segments": "Segments to edit when edit=true"
+      },
+      "performance": {
+        "generation_time": "~60-90 seconds on RTX 4090",
+        "vram_usage": "~16GB during generation",
+        "note": "Significantly faster than real-time music generation"
+      },
+      "notes": [
+        "This workflow uses DiffRhythm Full for 4 minute 45 second music generation",
+        "Best for complete song compositions with intro, development, and outro",
+        "All parameters except model and style_prompt are optional",
+        "Supports complex, multi-part compositions",
+        "Can optionally connect MultiLineLyricsDR node for lyrics input"
+      ]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-music/diffrhythm-reference-based-v1.json

@@ -0,0 +1,164 @@
+{
+  "last_node_id": 4,
+  "last_link_id": 3,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "LoadAudio",
+      "pos": [100, 100],
+      "size": [300, 100],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [1]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoadAudio"
+      },
+      "widgets_values": [
+        "reference_audio.wav"
+      ],
+      "title": "Load Reference Audio"
+    },
+    {
+      "id": 2,
+      "type": "DiffRhythmRun",
+      "pos": [500, 100],
+      "size": [400, 450],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "style_audio_or_edit_song",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [2, 3]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "DiffRhythmRun"
+      },
+      "widgets_values": [
+        "cfm_model_v1_2.pt",
+        "Energetic rock music with driving guitar riffs and powerful drums",
+        true,
+        "euler",
+        30,
+        5,
+        "speed",
+        456,
+        "randomize",
+        false,
+        "[-1, 20], [60, -1]"
+      ],
+      "title": "DiffRhythm Reference-Based Generation"
+    },
+    {
+      "id": 3,
+      "type": "PreviewAudio",
+      "pos": [1000, 100],
+      "size": [300, 100],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewAudio"
+      },
+      "title": "Preview Generated Audio"
+    },
+    {
+      "id": 4,
+      "type": "SaveAudio",
+      "pos": [1000, 250],
+      "size": [300, 100],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 3
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveAudio"
+      },
+      "widgets_values": [
+        "diffrhythm_reference_output"
+      ],
+      "title": "Save Audio"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"],
+    [2, 2, 0, 3, 0, "AUDIO"],
+    [3, 2, 0, 4, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "DiffRhythm Reference-Based Generation v1",
+      "description": "Generate new music based on a reference audio file while following text prompt guidance",
+      "version": "1.0.0",
+      "author": "valknar@pivoine.art",
+      "category": "text-to-music",
+      "tags": ["diffrhythm", "music-generation", "reference-based", "style-transfer"],
+      "requirements": {
+        "custom_nodes": ["ComfyUI_DiffRhythm"],
+        "models": ["ASLP-lab/DiffRhythm-1_2", "ASLP-lab/DiffRhythm-vae", "OpenMuQ/MuQ-MuLan-large", "OpenMuQ/MuQ-large-msd-iter", "FacebookAI/xlm-roberta-base"],
+        "vram_min": "14GB",
+        "vram_recommended": "18GB",
+        "system_deps": ["espeak-ng"]
+      },
+      "usage": {
+        "reference_audio": "Path to reference audio file (WAV, MP3, or other supported formats)",
+        "model": "cfm_model_v1_2.pt (DiffRhythm 1.2)",
+        "style_prompt": "Text description guiding the style and characteristics of generated music",
+        "unload_model": "Boolean to unload model after generation (default: true)",
+        "odeint_method": "ODE solver: euler, midpoint, rk4, implicit_adams (default: euler)",
+        "steps": "Number of diffusion steps: 1-100 (default: 30)",
+        "cfg": "Classifier-free guidance scale: 1-10 (default: 5 for reference-based)",
+        "quality_or_speed": "Generation mode: quality or speed (default: speed)",
+        "seed": "Random seed for reproducibility (default: 456)",
+        "edit": "Enable segment editing mode (default: false)",
+        "edit_segments": "Segments to edit when edit=true"
+      },
+      "use_cases": [
+        "Style transfer: Apply the style of reference music to new prompt",
+        "Variations: Create variations of existing compositions",
+        "Genre transformation: Transform music to different genre while keeping structure",
+        "Mood adaptation: Change the mood/emotion while maintaining musical elements"
+      ],
+      "notes": [
+        "This workflow combines reference audio with text prompt guidance",
+        "The reference audio is connected to the style_audio_or_edit_song input",
+        "Higher cfg values (7-10) = closer adherence to both prompt and reference",
+        "Lower cfg values (2-4) = more creative interpretation",
+        "Reference audio should ideally be similar duration to target (95s for cfm_model_v1_2.pt)",
+        "Can use any format supported by ComfyUI's LoadAudio node"
+      ]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-music/diffrhythm-simple-t2m-v1.json

@@ -0,0 +1,125 @@
+{
+  "last_node_id": 3,
+  "last_link_id": 2,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "DiffRhythmRun",
+      "pos": [100, 100],
+      "size": [400, 400],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [1, 2]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "DiffRhythmRun"
+      },
+      "widgets_values": [
+        "cfm_model_v1_2.pt",
+        "Upbeat electronic dance music with energetic beats and synthesizer melodies",
+        true,
+        "euler",
+        30,
+        4,
+        "speed",
+        42,
+        "randomize",
+        false,
+        "[-1, 20], [60, -1]"
+      ],
+      "title": "DiffRhythm Text-to-Music (95s)"
+    },
+    {
+      "id": 2,
+      "type": "PreviewAudio",
+      "pos": [600, 100],
+      "size": [300, 100],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewAudio"
+      },
+      "title": "Preview Audio"
+    },
+    {
+      "id": 3,
+      "type": "SaveAudio",
+      "pos": [600, 250],
+      "size": [300, 100],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveAudio"
+      },
+      "widgets_values": [
+        "diffrhythm_output"
+      ],
+      "title": "Save Audio"
+    }
+  ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"],
+    [2, 1, 0, 3, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "workflow_info": {
+      "name": "DiffRhythm Simple Text-to-Music v1",
+      "description": "Basic text-to-music generation using DiffRhythm 1.2 (95 seconds)",
+      "version": "1.0.0",
+      "author": "valknar@pivoine.art",
+      "category": "text-to-music",
+      "tags": ["diffrhythm", "music-generation", "text-to-music", "95s"],
+      "requirements": {
+        "custom_nodes": ["ComfyUI_DiffRhythm"],
+        "models": ["ASLP-lab/DiffRhythm-1_2", "ASLP-lab/DiffRhythm-vae", "OpenMuQ/MuQ-MuLan-large", "OpenMuQ/MuQ-large-msd-iter", "FacebookAI/xlm-roberta-base"],
+        "vram_min": "12GB",
+        "vram_recommended": "16GB",
+        "system_deps": ["espeak-ng"]
+      },
+      "usage": {
+        "model": "cfm_model_v1_2.pt (DiffRhythm 1.2 - 95s generation)",
+        "style_prompt": "Text description of the desired music style, mood, and instruments",
+        "unload_model": "Boolean to unload model after generation (default: true)",
+        "odeint_method": "ODE solver: euler, midpoint, rk4, implicit_adams (default: euler)",
+        "steps": "Number of diffusion steps: 1-100 (default: 30)",
+        "cfg": "Classifier-free guidance scale: 1-10 (default: 4)",
+        "quality_or_speed": "Generation mode: quality or speed (default: speed)",
+        "seed": "Random seed for reproducibility (default: 42)",
+        "edit": "Enable segment editing mode (default: false)",
+        "edit_segments": "Segments to edit when edit=true (default: [-1, 20], [60, -1])"
+      },
+      "notes": [
+        "This workflow uses DiffRhythm 1.2 for 95-second music generation",
+        "All parameters except model and style_prompt are optional",
+        "Supports English and Chinese text prompts",
+        "Generation time: ~30-60 seconds on RTX 4090",
+        "Can optionally connect MultiLineLyricsDR node for lyrics input"
+      ]
+    }
+  },
+  "version": 0.4
+}

comfyui/workflows/text-to-music/musicgen-large-t2m-production-v1.json

@@ -1,136 +1,151 @@
 {
-  "last_node_id": 6,
+  "last_node_id": 2,
+  "last_link_id": 1,
   "nodes": [
     {
       "id": 1,
-      "type": "MusicGenLoader",
+      "type": "Musicgen_",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [400, 300],
-        100
-      ],
-      "widgets_values": [
-        "facebook/musicgen-large"
-      ],
-      "title": "MusicGen Large Loader",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "links": [1],
+          "slot_index": 0
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenLoader"
+        "Node name for S&R": "Musicgen_"
       },
-      "size": {
+      "widgets_values": [
-        "0": 350,
+        "Upbeat electronic dance music with energetic beats and synthesizer melodies",
-        "1": 100
+        30.0,
-      }
+        4.0,
+        42,
+        "auto"
+      ],
+      "title": "MusicGen Large Generator"
     },
     {
       "id": 2,
-      "type": "MusicGenTextEncode",
+      "type": "AudioPlay",
-      "pos": [
+      "pos": [500, 100],
-        400,
+      "size": [315, 100],
-        100
-      ],
-      "widgets_values": [
-        "Upbeat electronic dance music"
-      ],
-      "title": "API Music Prompt",
       "flags": {},
       "order": 1,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenTextEncode"
+        "Node name for S&R": "AudioPlay"
       },
-      "size": {
+      "title": "API Audio Output"
-        "0": 400,
-        "1": 200
-      }
-    },
-    {
-      "id": 3,
-      "type": "MusicGenSampler",
-      "pos": [
-        750,
-        100
-      ],
-      "widgets_values": [
-        42,
-        "fixed",
-        30.0,
-        250,
-        3.0
-      ],
-      "title": "MusicGen Sampler",
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "MusicGenSampler"
-      },
-      "size": {
-        "0": 315,
-        "1": 474
-      }
-    },
-    {
-      "id": 4,
-      "type": "SaveAudio",
-      "pos": [
-        1100,
-        100
-      ],
-      "widgets_values": [
-        "musicgen_large_output",
-        "wav"
-      ],
-      "title": "API Audio Output",
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "AudioSave"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
-      "name": "MusicGen Large Text-to-Music",
+      "name": "MusicGen Large Text-to-Music Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
+      "description": "High-quality music generation using MusicGen Large. Generates up to 30 seconds of music from text prompts.",
       "category": "text-to-music",
-      "tags": [
+      "tags": ["musicgen", "large", "t2m", "audio", "production"],
-        "musicgen",
-        "large",
-        "t2m"
-      ],
       "requirements": {
-        "models": [
+        "models": ["facebook/musicgen-large"],
-          "musicgen-large"
+        "custom_nodes": ["comfyui-sound-lab"],
-        ],
+        "vram_min": "12GB",
-        "custom_nodes": [
+        "vram_recommended": "16GB"
-          "comfyui-sound-lab"
-        ],
-        "vram_min": "16GB"
       },
       "parameters": {
         "prompt": {
-          "node_id": 2,
+          "node_id": 1,
-          "type": "string"
+          "widget_index": 0,
+          "type": "string",
+          "required": true,
+          "default": "Upbeat electronic dance music",
+          "description": "Text description of desired music style, genre, mood, instruments"
         },
         "duration": {
-          "node_id": 3,
+          "node_id": 1,
-          "default": 30.0
+          "widget_index": 1,
+          "type": "float",
+          "required": false,
+          "default": 30.0,
+          "min": 1.0,
+          "max": 60.0,
+          "description": "Duration in seconds (up to 30s recommended for best quality)"
+        },
+        "guidance_scale": {
+          "node_id": 1,
+          "widget_index": 2,
+          "type": "float",
+          "required": false,
+          "default": 4.0,
+          "min": 0.0,
+          "max": 20.0,
+          "description": "How closely to follow the prompt (4.0 recommended)"
+        },
+        "seed": {
+          "node_id": 1,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 42,
+          "min": 0,
+          "max": 2147483647,
+          "description": "Random seed for reproducibility"
+        },
+        "device": {
+          "node_id": 1,
+          "widget_index": 4,
+          "type": "string",
+          "required": false,
+          "default": "auto",
+          "options": ["auto", "cpu"],
+          "description": "Computation device (auto recommended)"
+        }
+      },
+      "outputs": {
+        "audio": {
+          "node_id": 2,
+          "type": "audio",
+          "format": "FLAC",
+          "sample_rate": "32000 Hz",
+          "channels": "stereo"
         }
       },
       "performance": {
-        "avg_generation_time": "45-70 seconds",
+        "avg_generation_time": "60-90 seconds for 30s audio",
-        "vram_usage": "~14-16GB"
+        "vram_usage": "~14-16GB",
-      }
+        "gpu_utilization": "90-100%"
+      },
+      "use_cases": [
+        "Background music generation",
+        "Game soundtrack creation",
+        "Video background music",
+        "Music prototyping and ideation"
+      ],
+      "notes": [
+        "MusicGen Large produces highest quality output",
+        "Best results with clear, specific prompts",
+        "Can specify genre, mood, instruments, tempo",
+        "Output is 32kHz stereo FLAC format",
+        "Longer durations (>30s) may reduce coherence"
+      ]
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui/workflows/text-to-music/musicgen-medium-t2m-production-v1.json

@@ -1,136 +1,151 @@
 {
-  "last_node_id": 6,
+  "last_node_id": 2,
+  "last_link_id": 1,
   "nodes": [
     {
       "id": 1,
-      "type": "MusicGenLoader",
+      "type": "Musicgen_",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [400, 300],
-        100
-      ],
-      "widgets_values": [
-        "facebook/musicgen-medium"
-      ],
-      "title": "MusicGen Medium Loader",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "links": [1],
+          "slot_index": 0
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenLoader"
+        "Node name for S&R": "Musicgen_"
       },
-      "size": {
+      "widgets_values": [
-        "0": 350,
+        "Upbeat electronic dance music with energetic beats and synthesizer melodies",
-        "1": 100
+        30.0,
-      }
+        3.0,
+        42,
+        "auto"
+      ],
+      "title": "MusicGen Medium Generator"
     },
     {
       "id": 2,
-      "type": "MusicGenTextEncode",
+      "type": "AudioPlay",
-      "pos": [
+      "pos": [500, 100],
-        400,
+      "size": [315, 100],
-        100
-      ],
-      "widgets_values": [
-        "Upbeat electronic dance music"
-      ],
-      "title": "API Music Prompt",
       "flags": {},
       "order": 1,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenTextEncode"
+        "Node name for S&R": "AudioPlay"
       },
-      "size": {
+      "title": "API Audio Output"
-        "0": 400,
-        "1": 200
-      }
-    },
-    {
-      "id": 3,
-      "type": "MusicGenSampler",
-      "pos": [
-        750,
-        100
-      ],
-      "widgets_values": [
-        42,
-        "fixed",
-        30.0,
-        250,
-        3.0
-      ],
-      "title": "MusicGen Sampler",
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "MusicGenSampler"
-      },
-      "size": {
-        "0": 315,
-        "1": 474
-      }
-    },
-    {
-      "id": 4,
-      "type": "SaveAudio",
-      "pos": [
-        1100,
-        100
-      ],
-      "widgets_values": [
-        "musicgen_medium_output",
-        "wav"
-      ],
-      "title": "API Audio Output",
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "AudioSave"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
-      "name": "MusicGen Medium Text-to-Music",
+      "name": "MusicGen Medium Text-to-Music Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
+      "description": "Balanced music generation using MusicGen Medium. Good quality with moderate VRAM usage.",
       "category": "text-to-music",
-      "tags": [
+      "tags": ["musicgen", "medium", "t2m", "audio", "production"],
-        "musicgen",
-        "medium",
-        "t2m"
-      ],
       "requirements": {
-        "models": [
+        "models": ["facebook/musicgen-medium"],
-          "musicgen-medium"
+        "custom_nodes": ["comfyui-sound-lab"],
-        ],
+        "vram_min": "8GB",
-        "custom_nodes": [
+        "vram_recommended": "10GB"
-          "comfyui-sound-lab"
-        ],
-        "vram_min": "10GB"
       },
       "parameters": {
         "prompt": {
-          "node_id": 2,
+          "node_id": 1,
-          "type": "string"
+          "widget_index": 0,
+          "type": "string",
+          "required": true,
+          "default": "Upbeat electronic dance music",
+          "description": "Text description of desired music style, genre, mood, instruments"
         },
         "duration": {
-          "node_id": 3,
+          "node_id": 1,
-          "default": 30.0
+          "widget_index": 1,
+          "type": "float",
+          "required": false,
+          "default": 30.0,
+          "min": 1.0,
+          "max": 60.0,
+          "description": "Duration in seconds (up to 30s recommended for best quality)"
+        },
+        "guidance_scale": {
+          "node_id": 1,
+          "widget_index": 2,
+          "type": "float",
+          "required": false,
+          "default": 3.0,
+          "min": 0.0,
+          "max": 20.0,
+          "description": "How closely to follow the prompt (3.0 recommended)"
+        },
+        "seed": {
+          "node_id": 1,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 42,
+          "min": 0,
+          "max": 2147483647,
+          "description": "Random seed for reproducibility"
+        },
+        "device": {
+          "node_id": 1,
+          "widget_index": 4,
+          "type": "string",
+          "required": false,
+          "default": "auto",
+          "options": ["auto", "cpu"],
+          "description": "Computation device (auto recommended)"
+        }
+      },
+      "outputs": {
+        "audio": {
+          "node_id": 2,
+          "type": "audio",
+          "format": "FLAC",
+          "sample_rate": "32000 Hz",
+          "channels": "stereo"
         }
       },
       "performance": {
-        "avg_generation_time": "25-40 seconds",
+        "avg_generation_time": "30-50 seconds for 30s audio",
-        "vram_usage": "~8-10GB"
+        "vram_usage": "~8-10GB",
-      }
+        "gpu_utilization": "90-100%"
+      },
+      "use_cases": [
+        "Background music generation",
+        "Game soundtrack creation",
+        "Video background music",
+        "Music prototyping with moderate quality"
+      ],
+      "notes": [
+        "MusicGen Medium balances quality and resource usage",
+        "Faster than Large, better quality than Small",
+        "Best results with clear, specific prompts",
+        "Can specify genre, mood, instruments, tempo",
+        "Output is 32kHz stereo FLAC format"
+      ]
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui/workflows/text-to-music/musicgen-melody-t2m-production-v1.json

@@ -1,163 +1,150 @@
 {
-  "last_node_id": 7,
+  "last_node_id": 2,
+  "last_link_id": 1,
   "nodes": [
     {
       "id": 1,
-      "type": "LoadAudio",
+      "type": "Musicgen_",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [400, 300],
-        100
-      ],
-      "widgets_values": [
-        "melody_reference.wav"
-      ],
-      "title": "API Melody Reference",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "links": [1],
+          "slot_index": 0
+        }
+      ],
       "properties": {
-        "Node name for S&R": "LoadAudio"
+        "Node name for S&R": "Musicgen_"
       },
-      "size": {
+      "widgets_values": [
-        "0": 350,
+        "Electronic music with melodic elements and catchy hooks",
-        "1": 100
+        30.0,
-      }
+        3.5,
+        42,
+        "auto"
+      ],
+      "title": "MusicGen Melody Generator"
     },
     {
       "id": 2,
-      "type": "MusicGenLoader",
+      "type": "AudioPlay",
-      "pos": [
+      "pos": [500, 100],
-        50,
+      "size": [315, 100],
-        350
-      ],
-      "widgets_values": [
-        "facebook/musicgen-melody"
-      ],
-      "title": "MusicGen Melody Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenLoader"
+        "Node name for S&R": "AudioPlay"
       },
-      "size": {
+      "title": "API Audio Output"
-        "0": 350,
-        "1": 100
-      }
-    },
-    {
-      "id": 3,
-      "type": "MusicGenTextEncode",
-      "pos": [
-        400,
-        100
-      ],
-      "widgets_values": [
-        "Electronic music following the melody"
-      ],
-      "title": "API Music Prompt",
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "MusicGenTextEncode"
-      },
-      "size": {
-        "0": 400,
-        "1": 200
-      }
-    },
-    {
-      "id": 4,
-      "type": "MusicGenMelodySampler",
-      "pos": [
-        750,
-        100
-      ],
-      "widgets_values": [
-        42,
-        "fixed",
-        30.0,
-        250,
-        3.0
-      ],
-      "title": "MusicGen Melody Sampler",
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "MusicGenMelodySampler"
-      },
-      "size": {
-        "0": 315,
-        "1": 474
-      }
-    },
-    {
-      "id": 5,
-      "type": "SaveAudio",
-      "pos": [
-        1100,
-        100
-      ],
-      "widgets_values": [
-        "musicgen_melody_output",
-        "wav"
-      ],
-      "title": "API Audio Output",
-      "flags": {},
-      "order": 4,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "AudioSave"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
-      "name": "MusicGen Melody Text-to-Music",
+      "name": "MusicGen Melody Text-to-Music Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
+      "description": "Music generation using MusicGen with melodic focus. Note: Melody conditioning via audio input not yet supported.",
       "category": "text-to-music",
-      "tags": [
+      "tags": ["musicgen", "melody", "t2m", "audio", "production"],
-        "musicgen",
-        "melody",
-        "melody-conditioning",
-        "t2m"
-      ],
       "requirements": {
-        "models": [
+        "models": ["facebook/musicgen-melody"],
-          "musicgen-melody"
+        "custom_nodes": ["comfyui-sound-lab"],
-        ],
+        "vram_min": "10GB",
-        "custom_nodes": [
+        "vram_recommended": "12GB"
-          "comfyui-sound-lab"
-        ],
-        "vram_min": "12GB"
       },
       "parameters": {
-        "melody_audio": {
-          "node_id": 1,
-          "type": "audio"
-        },
         "prompt": {
-          "node_id": 3,
+          "node_id": 1,
-          "type": "string"
+          "widget_index": 0,
+          "type": "string",
+          "required": true,
+          "default": "Electronic music with melodic elements",
+          "description": "Text description of desired music style, genre, mood, instruments"
         },
         "duration": {
-          "node_id": 4,
+          "node_id": 1,
-          "default": 30.0
+          "widget_index": 1,
+          "type": "float",
+          "required": false,
+          "default": 30.0,
+          "min": 1.0,
+          "max": 60.0,
+          "description": "Duration in seconds (up to 30s recommended for best quality)"
+        },
+        "guidance_scale": {
+          "node_id": 1,
+          "widget_index": 2,
+          "type": "float",
+          "required": false,
+          "default": 3.5,
+          "min": 0.0,
+          "max": 20.0,
+          "description": "How closely to follow the prompt (3.5 recommended)"
+        },
+        "seed": {
+          "node_id": 1,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 42,
+          "min": 0,
+          "max": 2147483647,
+          "description": "Random seed for reproducibility"
+        },
+        "device": {
+          "node_id": 1,
+          "widget_index": 4,
+          "type": "string",
+          "required": false,
+          "default": "auto",
+          "options": ["auto", "cpu"],
+          "description": "Computation device (auto recommended)"
+        }
+      },
+      "outputs": {
+        "audio": {
+          "node_id": 2,
+          "type": "audio",
+          "format": "FLAC",
+          "sample_rate": "32000 Hz",
+          "channels": "stereo"
         }
       },
       "performance": {
-        "avg_generation_time": "35-55 seconds",
+        "avg_generation_time": "40-60 seconds for 30s audio",
-        "vram_usage": "~10-12GB"
+        "vram_usage": "~10-12GB",
-      }
+        "gpu_utilization": "90-100%"
+      },
+      "use_cases": [
+        "Melodic music generation",
+        "Creating catchy hooks and melodies",
+        "Music with strong melodic structure"
+      ],
+      "notes": [
+        "IMPORTANT: Melody conditioning via audio input not yet supported in current Sound Lab implementation",
+        "This workflow uses text prompts only - describe melodic characteristics in the prompt",
+        "For best melodic results, use descriptive terms like 'catchy melody', 'melodic hooks', 'harmonic progression'",
+        "MusicGen Melody model trained for melodic content",
+        "Output is 32kHz stereo FLAC format"
+      ]
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui/workflows/text-to-music/musicgen-small-t2m-production-v1.json

@@ -1,136 +1,151 @@
 {
-  "last_node_id": 6,
+  "last_node_id": 2,
+  "last_link_id": 1,
   "nodes": [
     {
       "id": 1,
-      "type": "MusicGenLoader",
+      "type": "Musicgen_",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [400, 300],
-        100
-      ],
-      "widgets_values": [
-        "facebook/musicgen-small"
-      ],
-      "title": "MusicGen Small Loader",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "links": [1],
+          "slot_index": 0
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenLoader"
+        "Node name for S&R": "Musicgen_"
       },
-      "size": {
+      "widgets_values": [
-        "0": 350,
+        "Upbeat electronic dance music with energetic beats and synthesizer melodies",
-        "1": 100
+        30.0,
-      }
+        3.0,
+        42,
+        "auto"
+      ],
+      "title": "MusicGen Small Generator"
     },
     {
       "id": 2,
-      "type": "MusicGenTextEncode",
+      "type": "AudioPlay",
-      "pos": [
+      "pos": [500, 100],
-        400,
+      "size": [315, 100],
-        100
-      ],
-      "widgets_values": [
-        "Upbeat electronic dance music"
-      ],
-      "title": "API Music Prompt",
       "flags": {},
       "order": 1,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 1
+        }
+      ],
       "properties": {
-        "Node name for S&R": "MusicGenTextEncode"
+        "Node name for S&R": "AudioPlay"
       },
-      "size": {
+      "title": "API Audio Output"
-        "0": 400,
-        "1": 200
-      }
-    },
-    {
-      "id": 3,
-      "type": "MusicGenSampler",
-      "pos": [
-        750,
-        100
-      ],
-      "widgets_values": [
-        42,
-        "fixed",
-        30.0,
-        250,
-        3.0
-      ],
-      "title": "MusicGen Sampler",
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "MusicGenSampler"
-      },
-      "size": {
-        "0": 315,
-        "1": 474
-      }
-    },
-    {
-      "id": 4,
-      "type": "SaveAudio",
-      "pos": [
-        1100,
-        100
-      ],
-      "widgets_values": [
-        "musicgen_small_output",
-        "wav"
-      ],
-      "title": "API Audio Output",
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "AudioSave"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 2, 0, "AUDIO"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
-      "name": "MusicGen Small Text-to-Music",
+      "name": "MusicGen Small Text-to-Music Production",
-      "version": "1.0.0",
+      "version": "1.2.0",
+      "description": "Fast music generation using MusicGen Small. Lower quality but fastest generation and minimal VRAM.",
       "category": "text-to-music",
-      "tags": [
+      "tags": ["musicgen", "small", "t2m", "audio", "production", "fast"],
-        "musicgen",
-        "small",
-        "t2m"
-      ],
       "requirements": {
-        "models": [
+        "models": ["facebook/musicgen-small"],
-          "musicgen-small"
+        "custom_nodes": ["comfyui-sound-lab"],
-        ],
+        "vram_min": "4GB",
-        "custom_nodes": [
+        "vram_recommended": "6GB"
-          "comfyui-sound-lab"
-        ],
-        "vram_min": "6GB"
       },
       "parameters": {
         "prompt": {
-          "node_id": 2,
+          "node_id": 1,
-          "type": "string"
+          "widget_index": 0,
+          "type": "string",
+          "required": true,
+          "default": "Upbeat electronic dance music",
+          "description": "Text description of desired music style, genre, mood, instruments"
         },
         "duration": {
-          "node_id": 3,
+          "node_id": 1,
-          "default": 30.0
+          "widget_index": 1,
+          "type": "float",
+          "required": false,
+          "default": 30.0,
+          "min": 1.0,
+          "max": 60.0,
+          "description": "Duration in seconds (up to 30s recommended for best quality)"
+        },
+        "guidance_scale": {
+          "node_id": 1,
+          "widget_index": 2,
+          "type": "float",
+          "required": false,
+          "default": 3.0,
+          "min": 0.0,
+          "max": 20.0,
+          "description": "How closely to follow the prompt (3.0 recommended)"
+        },
+        "seed": {
+          "node_id": 1,
+          "widget_index": 3,
+          "type": "integer",
+          "required": false,
+          "default": 42,
+          "min": 0,
+          "max": 2147483647,
+          "description": "Random seed for reproducibility"
+        },
+        "device": {
+          "node_id": 1,
+          "widget_index": 4,
+          "type": "string",
+          "required": false,
+          "default": "auto",
+          "options": ["auto", "cpu"],
+          "description": "Computation device (auto recommended)"
+        }
+      },
+      "outputs": {
+        "audio": {
+          "node_id": 2,
+          "type": "audio",
+          "format": "FLAC",
+          "sample_rate": "32000 Hz",
+          "channels": "stereo"
         }
       },
       "performance": {
-        "avg_generation_time": "15-25 seconds",
+        "avg_generation_time": "20-35 seconds for 30s audio",
-        "vram_usage": "~4-6GB"
+        "vram_usage": "~4-6GB",
-      }
+        "gpu_utilization": "85-95%"
+      },
+      "use_cases": [
+        "Rapid prototyping",
+        "Low-resource environments",
+        "Quick background music drafts",
+        "Testing different prompts quickly"
+      ],
+      "notes": [
+        "MusicGen Small is fastest but lowest quality",
+        "Good for rapid iteration and testing",
+        "Best results with clear, specific prompts",
+        "Can specify genre, mood, instruments, tempo",
+        "Output is 32kHz stereo FLAC format"
+      ]
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui/workflows/upscaling/face-upscale-production-v1.json

@@ -1,222 +1,185 @@
 {
-  "last_node_id": 10,
+  "last_node_id": 5,
+  "last_link_id": 4,
   "nodes": [
     {
       "id": 1,
       "type": "LoadImage",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [350, 100],
-        100
-      ],
-      "widgets_values": [
-        "input_portrait.png",
-        "image"
-      ],
-      "title": "API Input Portrait",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null,
+          "slot_index": 1
+        }
+      ],
       "properties": {
         "Node name for S&R": "LoadImage"
       },
-      "size": {
+      "widgets_values": ["input_portrait.png", "image"],
-        "0": 350,
+      "title": "API Input Portrait"
-        "1": 100
-      }
     },
     {
       "id": 2,
-      "type": "CheckpointLoaderSimple",
+      "type": "UpscaleModelLoader",
-      "pos": [
+      "pos": [50, 400],
-        50,
+      "size": [350, 100],
-        400
-      ],
-      "widgets_values": [
-        "diffusers/stable-diffusion-xl-base-1.0"
-      ],
-      "title": "SDXL Checkpoint Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
-      "properties": {
+      "outputs": [
-        "Node name for S&R": "CheckpointLoaderSimple"
+        {
-      },
+          "name": "UPSCALE_MODEL",
-      "size": {
+          "type": "UPSCALE_MODEL",
-        "0": 350,
+          "links": [2],
-        "1": 100
+          "slot_index": 0
-      }
+        }
-    },
-    {
-      "id": 3,
-      "type": "FaceDetailer",
-      "pos": [
-        450,
-        100
       ],
-      "widgets_values": [
-        "bbox/face_yolov8m.pt",
-        512,
-        0.5,
-        20,
-        8.0,
-        0.35,
-        10
-      ],
-      "title": "Face Detailer",
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "FaceDetailer"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
-    },
-    {
-      "id": 4,
-      "type": "UpscaleModelLoader",
-      "pos": [
-        50,
-        650
-      ],
-      "widgets_values": [
-        "RealESRGAN_x2.pth"
-      ],
-      "title": "Face Upscale Model",
-      "flags": {},
-      "order": 3,
-      "mode": 0,
       "properties": {
         "Node name for S&R": "UpscaleModelLoader"
       },
-      "size": {
+      "widgets_values": ["RealESRGAN_x2.pth"],
-        "0": 350,
+      "title": "Face Upscale Model (2x)"
-        "1": 100
-      }
     },
     {
-      "id": 5,
+      "id": 3,
       "type": "ImageUpscaleWithModel",
-      "pos": [
+      "pos": [450, 100],
-        800,
+      "size": [315, 100],
-        100
-      ],
-      "title": "Upscale Face Regions",
       "flags": {},
-      "order": 4,
+      "order": 2,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "upscale_model",
+          "type": "UPSCALE_MODEL",
+          "link": 2
+        },
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [3, 4],
+          "slot_index": 0
+        }
+      ],
       "properties": {
         "Node name for S&R": "ImageUpscaleWithModel"
       },
-      "size": {
+      "title": "Upscale Portrait (2x)"
-        "0": 315,
-        "1": 100
-      }
     },
     {
-      "id": 6,
+      "id": 4,
       "type": "PreviewImage",
-      "pos": [
+      "pos": [800, 100],
-        1150,
+      "size": [315, 100],
-        100
-      ],
-      "title": "Preview Enhanced",
       "flags": {},
-      "order": 5,
+      "order": 3,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 3
+        }
+      ],
       "properties": {
         "Node name for S&R": "PreviewImage"
       },
-      "size": {
+      "title": "Preview Enhanced"
-        "0": 315,
-        "1": 100
-      }
     },
     {
-      "id": 7,
+      "id": 5,
       "type": "SaveImage",
-      "pos": [
+      "pos": [800, 400],
-        1150,
+      "size": [315, 100],
-        400
-      ],
-      "widgets_values": [
-        "face_upscale_output"
-      ],
-      "title": "API Image Output",
       "flags": {},
-      "order": 6,
+      "order": 4,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 4
+        }
+      ],
       "properties": {
         "Node name for S&R": "SaveImage"
       },
-      "size": {
+      "widgets_values": ["face_upscale_output"],
-        "0": 315,
+      "title": "API Image Output"
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 3, 1, "IMAGE"],
+    [2, 2, 0, 3, 0, "UPSCALE_MODEL"],
+    [3, 3, 0, 4, 0, "IMAGE"],
+    [4, 3, 0, 5, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
       "name": "Face Upscale Production",
-      "version": "1.0.0",
+      "version": "1.1.0",
-      "description": "Portrait-focused upscaling with FaceDetailer from Impact-Pack. Detects and enhances faces with special attention to facial details.",
+      "description": "Portrait-focused upscaling using RealESRGAN x2. Simplified workflow optimized for face enhancement. For advanced face detection with FaceDetailer, additional nodes are required.",
       "category": "upscaling",
-      "tags": [
+      "tags": ["face-upscale", "portrait", "realesrgan", "production"],
-        "face-upscale",
-        "portrait",
-        "facedetailer",
-        "impact-pack",
-        "production"
-      ],
       "requirements": {
-        "models": [
+        "models": ["RealESRGAN"],
-          "stable-diffusion-xl-base-1.0",
+        "custom_nodes": [],
-          "face_yolov8m",
+        "vram_min": "8GB"
-          "RealESRGAN"
-        ],
-        "custom_nodes": [
-          "ComfyUI-Impact-Pack"
-        ],
-        "vram_min": "14GB"
       },
       "parameters": {
         "input_image": {
           "node_id": 1,
           "type": "image",
-          "required": true
+          "required": true,
+          "description": "Portrait image to upscale"
         },
-        "detection_threshold": {
+        "model": {
-          "node_id": 3,
+          "node_id": 2,
-          "type": "float",
+          "widget_index": 0,
-          "default": 0.5,
+          "type": "string",
-          "description": "Face detection confidence"
+          "default": "RealESRGAN_x2.pth",
-        },
+          "options": ["RealESRGAN_x2.pth", "RealESRGAN_x4.pth"],
-        "detail_steps": {
+          "description": "Upscale model (x2 recommended for portraits)"
-          "node_id": 3,
-          "type": "integer",
-          "default": 20
-        },
-        "denoise": {
-          "node_id": 3,
-          "type": "float",
-          "default": 0.35
         }
       },
       "performance": {
-        "avg_generation_time": "30-50 seconds",
+        "avg_generation_time": "5-15 seconds",
-        "vram_usage": "~12-16GB"
+        "vram_usage": "~6-8GB"
       },
       "use_cases": [
         "Portrait enhancement",
         "Professional headshots",
         "Face restoration",
         "ID photo upscaling"
+      ],
+      "notes": [
+        "This is a simplified version using basic upscaling",
+        "For advanced face detection, use FaceDetailer from Impact-Pack",
+        "FaceDetailer requires: BBOX detector, SDXL checkpoint, CLIP conditioning",
+        "Current version optimized for speed and simplicity"
       ]
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui/workflows/upscaling/simple-upscale-production-v1.json

@@ -1,151 +1,179 @@
 {
   "last_node_id": 6,
+  "last_link_id": 5,
   "nodes": [
     {
       "id": 1,
       "type": "LoadImage",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [350, 100],
-        100
-      ],
-      "widgets_values": [
-        "input_image.png",
-        "image"
-      ],
-      "title": "API Input Image",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null,
+          "slot_index": 1
+        }
+      ],
       "properties": {
         "Node name for S&R": "LoadImage"
       },
-      "size": {
+      "widgets_values": ["input_image.png", "image"],
-        "0": 350,
+      "title": "API Input Image"
-        "1": 100
-      }
     },
     {
       "id": 2,
       "type": "UpscaleModelLoader",
-      "pos": [
+      "pos": [50, 400],
-        50,
+      "size": [350, 100],
-        400
-      ],
-      "widgets_values": [
-        "RealESRGAN_x4.pth"
-      ],
-      "title": "Upscale Model Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "UPSCALE_MODEL",
+          "type": "UPSCALE_MODEL",
+          "links": [2],
+          "slot_index": 0
+        }
+      ],
       "properties": {
         "Node name for S&R": "UpscaleModelLoader"
       },
-      "size": {
+      "widgets_values": ["RealESRGAN_x4.pth"],
-        "0": 350,
+      "title": "Upscale Model Loader"
-        "1": 100
-      }
     },
     {
       "id": 3,
       "type": "ImageUpscaleWithModel",
-      "pos": [
+      "pos": [450, 100],
-        450,
+      "size": [315, 100],
-        100
-      ],
-      "title": "Upscale with Model (4x)",
       "flags": {},
       "order": 2,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "upscale_model",
+          "type": "UPSCALE_MODEL",
+          "link": 2
+        },
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [3],
+          "slot_index": 0
+        }
+      ],
       "properties": {
         "Node name for S&R": "ImageUpscaleWithModel"
       },
-      "size": {
+      "title": "Upscale with Model (4x)"
-        "0": 315,
-        "1": 100
-      }
     },
     {
       "id": 4,
-      "type": "ImageScale",
+      "type": "ImageScaleBy",
-      "pos": [
+      "pos": [800, 100],
-        800,
+      "size": [315, 100],
-        100
-      ],
-      "widgets_values": [
-        "lanczos",
-        2
-      ],
-      "title": "Optional Downscale to 2x",
       "flags": {},
       "order": 3,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [4, 5],
+          "slot_index": 0
+        }
+      ],
       "properties": {
-        "Node name for S&R": "ImageScale"
+        "Node name for S&R": "ImageScaleBy"
       },
-      "size": {
+      "widgets_values": ["lanczos", 0.5],
-        "0": 315,
+      "title": "Optional Downscale to 2x"
-        "1": 100
-      }
     },
     {
       "id": 5,
       "type": "PreviewImage",
-      "pos": [
+      "pos": [1150, 100],
-        1150,
+      "size": [315, 100],
-        100
-      ],
-      "title": "Preview Output",
       "flags": {},
       "order": 4,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 4
+        }
+      ],
       "properties": {
         "Node name for S&R": "PreviewImage"
       },
-      "size": {
+      "title": "Preview Output"
-        "0": 315,
-        "1": 100
-      }
     },
     {
       "id": 6,
       "type": "SaveImage",
-      "pos": [
+      "pos": [1150, 400],
-        1150,
+      "size": [315, 100],
-        400
-      ],
-      "widgets_values": [
-        "simple_upscale_output"
-      ],
-      "title": "API Image Output",
       "flags": {},
       "order": 5,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 5
+        }
+      ],
       "properties": {
         "Node name for S&R": "SaveImage"
       },
-      "size": {
+      "widgets_values": ["simple_upscale_output"],
-        "0": 315,
+      "title": "API Image Output"
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 3, 1, "IMAGE"],
+    [2, 2, 0, 3, 0, "UPSCALE_MODEL"],
+    [3, 3, 0, 4, 0, "IMAGE"],
+    [4, 4, 0, 5, 0, "IMAGE"],
+    [5, 4, 0, 6, 0, "IMAGE"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
       "name": "Simple Upscale Production",
       "version": "1.0.0",
       "description": "Fast, straightforward upscaling using RealESRGAN. No diffusion refinement, optimized for speed.",
       "category": "upscaling",
-      "tags": [
+      "tags": ["simple-upscale", "fast", "realesrgan", "production"],
-        "simple-upscale",
-        "fast",
-        "realesrgan",
-        "production"
-      ],
       "requirements": {
-        "models": [
+        "models": ["RealESRGAN"],
-          "RealESRGAN"
-        ],
         "custom_nodes": [],
         "vram_min": "8GB"
       },
@@ -159,15 +187,14 @@
           "node_id": 2,
           "type": "string",
           "default": "RealESRGAN_x4.pth",
-          "options": [
+          "options": ["RealESRGAN_x2.pth", "RealESRGAN_x4.pth"]
-            "RealESRGAN_x2.pth",
-            "RealESRGAN_x4.pth"
-          ]
         },
         "final_scale": {
           "node_id": 4,
-          "type": "integer",
+          "widget_index": 1,
-          "default": 2
+          "type": "float",
+          "default": 0.5,
+          "description": "Scale factor after 4x upscale (0.5 = final 2x, 1.0 = keep 4x)"
         }
       },
       "performance": {
@@ -176,7 +203,5 @@
       }
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui/workflows/upscaling/ultimate-sd-upscale-production-v1.json

@@ -1,227 +1,269 @@
 {
-  "last_node_id": 12,
+  "last_node_id": 8,
+  "last_link_id": 10,
   "nodes": [
     {
       "id": 1,
       "type": "LoadImage",
-      "pos": [
+      "pos": [50, 100],
-        50,
+      "size": [350, 100],
-        100
-      ],
-      "widgets_values": [
-        "input_image.png",
-        "image"
-      ],
-      "title": "API Input Image",
       "flags": {},
       "order": 0,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [1],
+          "slot_index": 0
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null,
+          "slot_index": 1
+        }
+      ],
       "properties": {
         "Node name for S&R": "LoadImage"
       },
-      "size": {
+      "widgets_values": ["input_image.png", "image"],
-        "0": 350,
+      "title": "API Input Image"
-        "1": 100
-      }
     },
     {
       "id": 2,
       "type": "CheckpointLoaderSimple",
-      "pos": [
+      "pos": [50, 250],
-        50,
+      "size": [350, 100],
-        400
-      ],
-      "widgets_values": [
-        "diffusers/stable-diffusion-xl-base-1.0"
-      ],
-      "title": "SDXL Checkpoint Loader",
       "flags": {},
       "order": 1,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [2],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [3, 4],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [5],
+          "slot_index": 2
+        }
+      ],
       "properties": {
         "Node name for S&R": "CheckpointLoaderSimple"
       },
-      "size": {
+      "widgets_values": ["sd_xl_base_1.0.safetensors"],
-        "0": 350,
+      "title": "SDXL Checkpoint Loader"
-        "1": 100
-      }
     },
     {
       "id": 3,
-      "type": "UpscaleModelLoader",
+      "type": "CLIPTextEncode",
-      "pos": [
+      "pos": [450, 100],
-        50,
+      "size": [400, 200],
-        600
-      ],
-      "widgets_values": [
-        "RealESRGAN_x2.pth"
-      ],
-      "title": "Upscale Model Loader",
       "flags": {},
       "order": 2,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [6],
+          "slot_index": 0
+        }
+      ],
       "properties": {
-        "Node name for S&R": "UpscaleModelLoader"
+        "Node name for S&R": "CLIPTextEncode"
       },
-      "size": {
+      "widgets_values": ["high quality, detailed, sharp"],
-        "0": 350,
+      "title": "API Positive Prompt"
-        "1": 100
-      }
     },
     {
       "id": 4,
       "type": "CLIPTextEncode",
-      "pos": [
+      "pos": [450, 350],
-        450,
+      "size": [400, 200],
-        100
-      ],
-      "widgets_values": [
-        "high quality, detailed, sharp"
-      ],
-      "title": "API Positive Prompt",
       "flags": {},
       "order": 3,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [7],
+          "slot_index": 0
+        }
+      ],
       "properties": {
         "Node name for S&R": "CLIPTextEncode"
       },
-      "size": {
+      "widgets_values": ["blurry, low quality"],
-        "0": 400,
+      "title": "API Negative Prompt"
-        "1": 200
-      }
-    },
-    {
-      "id": 5,
-      "type": "CLIPTextEncode",
-      "pos": [
-        450,
-        300
-      ],
-      "widgets_values": [
-        "blurry, low quality"
-      ],
-      "title": "API Negative Prompt",
-      "flags": {},
-      "order": 4,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "CLIPTextEncode"
-      },
-      "size": {
-        "0": 400,
-        "1": 200
-      }
-    },
-    {
-      "id": 6,
-      "type": "UltimateSDUpscale",
-      "pos": [
-        800,
-        100
-      ],
-      "widgets_values": [
-        2,
-        42,
-        20,
-        8.0,
-        "dpmpp_2m",
-        "karras",
-        0.3,
-        "Linear",
-        512,
-        64
-      ],
-      "title": "Ultimate SD Upscale (2x)",
-      "flags": {},
-      "order": 5,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "UltimateSDUpscale"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
-    },
-    {
-      "id": 7,
-      "type": "VAEDecode",
-      "pos": [
-        1150,
-        100
-      ],
-      "title": "VAE Decode",
-      "flags": {},
-      "order": 6,
-      "mode": 0,
-      "properties": {
-        "Node name for S&R": "VAEDecode"
-      },
-      "size": {
-        "0": 315,
-        "1": 100
-      }
     },
     {
       "id": 8,
-      "type": "PreviewImage",
+      "type": "UpscaleModelLoader",
-      "pos": [
+      "pos": [50, 400],
-        1450,
+      "size": [350, 100],
-        100
-      ],
-      "title": "Preview Output",
       "flags": {},
-      "order": 7,
+      "order": 4,
       "mode": 0,
+      "outputs": [
+        {
+          "name": "UPSCALE_MODEL",
+          "type": "UPSCALE_MODEL",
+          "links": [10],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "UpscaleModelLoader"
+      },
+      "widgets_values": ["RealESRGAN_x2.pth"],
+      "title": "Upscale Model Loader"
+    },
+    {
+      "id": 5,
+      "type": "UltimateSDUpscale",
+      "pos": [900, 100],
+      "size": [315, 100],
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 1
+        },
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 2
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 6
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 7
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 5
+        },
+        {
+          "name": "upscale_model",
+          "type": "UPSCALE_MODEL",
+          "link": 10
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [8, 9],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "UltimateSDUpscale"
+      },
+      "widgets_values": [2, 42, "randomize", 20, 8.0, "dpmpp_2m", "karras", 0.3, "Linear", 512, 512, 8, 32, "None", 1.0, 64, 8, 16, true, false],
+      "title": "Ultimate SD Upscale (2x)"
+    },
+    {
+      "id": 6,
+      "type": "PreviewImage",
+      "pos": [1270, 100],
+      "size": [315, 100],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 8
+        }
+      ],
       "properties": {
         "Node name for S&R": "PreviewImage"
       },
-      "size": {
+      "title": "Preview Output"
-        "0": 315,
-        "1": 100
-      }
     },
     {
-      "id": 9,
+      "id": 7,
       "type": "SaveImage",
-      "pos": [
+      "pos": [1270, 400],
-        1450,
+      "size": [315, 100],
-        550
-      ],
-      "widgets_values": [
-        "ultimate_upscale_output"
-      ],
-      "title": "API Image Output",
       "flags": {},
-      "order": 8,
+      "order": 7,
       "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 9
+        }
+      ],
       "properties": {
         "Node name for S&R": "SaveImage"
       },
-      "size": {
+      "widgets_values": ["ultimate_upscale_output"],
-        "0": 315,
+      "title": "API Image Output"
-        "1": 100
-      }
     }
   ],
+  "links": [
+    [1, 1, 0, 5, 0, "IMAGE"],
+    [2, 2, 0, 5, 1, "MODEL"],
+    [3, 2, 1, 3, 0, "CLIP"],
+    [4, 2, 1, 4, 0, "CLIP"],
+    [5, 2, 2, 5, 4, "VAE"],
+    [6, 3, 0, 5, 2, "CONDITIONING"],
+    [7, 4, 0, 5, 3, "CONDITIONING"],
+    [8, 5, 0, 6, 0, "IMAGE"],
+    [9, 5, 0, 7, 0, "IMAGE"],
+    [10, 8, 0, 5, 5, "UPSCALE_MODEL"]
+  ],
+  "groups": [],
+  "config": {},
   "extra": {
     "workflow_info": {
       "name": "Ultimate SD Upscale Production",
-      "version": "1.0.0",
+      "version": "1.1.0",
       "description": "Professional upscaling with Ultimate SD Upscale. Combines AI upscaling with diffusion refinement for superior detail and quality.",
       "category": "upscaling",
-      "tags": [
+      "tags": ["ultimate-sd-upscale", "upscaling", "enhancement", "production"],
-        "ultimate-sd-upscale",
-        "upscaling",
-        "enhancement",
-        "production"
-      ],
       "requirements": {
-        "models": [
+        "models": ["stable-diffusion-xl-base-1.0", "RealESRGAN"],
-          "stable-diffusion-xl-base-1.0",
+        "custom_nodes": ["ComfyUI_UltimateSDUpscale"],
-          "RealESRGAN"
-        ],
-        "custom_nodes": [],
         "vram_min": "18GB"
       },
       "parameters": {
@@ -230,26 +272,57 @@
           "type": "image",
           "required": true
         },
-        "scale": {
+        "positive_prompt": {
-          "node_id": 6,
+          "node_id": 3,
-          "type": "integer",
+          "widget_index": 0,
-          "default": 2,
+          "type": "string",
-          "options": [
+          "default": "high quality, detailed, sharp",
-            2,
+          "description": "Enhancement prompt"
-            4
-          ]
         },
-        "tile_size": {
+        "negative_prompt": {
-          "node_id": 6,
+          "node_id": 4,
-          "type": "integer",
+          "widget_index": 0,
-          "default": 512,
+          "type": "string",
-          "description": "Processing tile size"
+          "default": "blurry, low quality",
+          "description": "Qualities to avoid"
+        },
+        "upscale_model": {
+          "node_id": 8,
+          "widget_index": 0,
+          "type": "string",
+          "default": "RealESRGAN_x2.pth",
+          "options": ["RealESRGAN_x2.pth", "RealESRGAN_x4.pth"],
+          "description": "Upscale model to use"
+        },
+        "upscale_by": {
+          "node_id": 5,
+          "widget_index": 0,
+          "type": "float",
+          "default": 2,
+          "min": 0.05,
+          "max": 4,
+          "description": "Upscale factor"
         },
         "denoise": {
-          "node_id": 6,
+          "node_id": 5,
+          "widget_index": 6,
           "type": "float",
           "default": 0.3,
           "description": "Refinement strength"
+        },
+        "tile_width": {
+          "node_id": 5,
+          "widget_index": 8,
+          "type": "integer",
+          "default": 512,
+          "description": "Tile width for processing"
+        },
+        "tile_height": {
+          "node_id": 5,
+          "widget_index": 9,
+          "type": "integer",
+          "default": 512,
+          "description": "Tile height for processing"
         }
       },
       "performance": {
@@ -258,7 +331,5 @@
       }
     }
   },
-  "version": 0.4,
+  "version": 0.4
-  "links": [],
+}
-  "last_link_id": 0
-}

comfyui_models.yaml

@@ -1,471 +0,0 @@
-# ============================================================================
-# ComfyUI Model Configuration
-# ============================================================================
-#
-# This configuration file defines all available ComfyUI models for download.
-# Models are organized by category: image, video, audio, and support models.
-#
-# Each model entry contains:
-#   - repo_id: HuggingFace repository identifier
-#   - description: Human-readable description
-#   - size_gb: Approximate size in gigabytes
-#   - essential: Whether this is an essential model (true/false)
-#   - category: Model category (image/video/audio/support)
-#
-# ============================================================================
-
-# Global settings
-settings:
-  cache_dir: /workspace/huggingface_cache
-  parallel_downloads: 1
-  retry_attempts: 3
-  timeout_seconds: 3600
-
-# Model categories
-model_categories:
-  # ==========================================================================
-  # IMAGE GENERATION MODELS
-  # ==========================================================================
-  image_models:
-    - repo_id: black-forest-labs/FLUX.1-schnell
-      description: FLUX.1 Schnell - Fast 4-step inference
-      size_gb: 23
-      essential: true
-      category: image
-      type: diffusion_models
-      format: fp16
-      vram_gb: 23
-      notes: Industry-leading image generation quality
-      files:
-        - source: "flux1-schnell.safetensors"
-          dest: "flux1-schnell.safetensors"
-
-    - repo_id: black-forest-labs/FLUX.1-dev
-      description: FLUX.1 Dev - Balanced quality/speed
-      size_gb: 23
-      essential: false
-      category: image
-      type: diffusion_models
-      format: fp16
-      vram_gb: 23
-      notes: Development version with enhanced features
-      files:
-        - source: "flux1-dev.safetensors"
-          dest: "flux1-dev.safetensors"
-
-    - repo_id: runwayml/stable-diffusion-v1-5
-      description: SD 1.5 - For AnimateDiff
-      size_gb: 4
-      essential: true
-      category: image
-      type: checkpoints
-      format: fp16
-      vram_gb: 8
-      notes: Stable Diffusion 1.5 required for AnimateDiff motion modules
-      files:
-        - source: "v1-5-pruned-emaonly.safetensors"
-          dest: "v1-5-pruned-emaonly.safetensors"
-
-    - repo_id: stabilityai/stable-diffusion-xl-base-1.0
-      description: SDXL Base 1.0 - Industry standard
-      size_gb: 7
-      essential: true
-      category: image
-      type: checkpoints
-      format: fp16
-      vram_gb: 12
-      notes: Most widely used Stable Diffusion model
-      files:
-        - source: "sd_xl_base_1.0.safetensors"
-          dest: "sd_xl_base_1.0.safetensors"
-
-    - repo_id: stabilityai/stable-diffusion-xl-refiner-1.0
-      description: SDXL Refiner 1.0 - Enhances base output
-      size_gb: 6
-      essential: false
-      category: image
-      type: checkpoints
-      format: fp16
-      vram_gb: 12
-      notes: Use after SDXL base for improved details
-      files:
-        - source: "sd_xl_refiner_1.0.safetensors"
-          dest: "sd_xl_refiner_1.0.safetensors"
-
-    - repo_id: stabilityai/stable-diffusion-3.5-large
-      description: SD 3.5 Large - Latest Stability AI
-      size_gb: 18
-      essential: false
-      category: image
-      type: checkpoints
-      format: fp16
-      vram_gb: 20
-      notes: Newest generation Stable Diffusion
-      files:
-        - source: "sd3.5_large.safetensors"
-          dest: "sd3.5_large.safetensors"
-
-  # ==========================================================================
-  # VIDEO GENERATION MODELS
-  # ==========================================================================
-  video_models:
-    - repo_id: THUDM/CogVideoX-5b
-      description: CogVideoX-5B - Professional text-to-video
-      size_gb: 20
-      essential: true
-      category: video
-      type: diffusion_models
-      format: fp16
-      vram_gb: 20
-      frames: 49
-      resolution: 720p
-      notes: State-of-the-art text-to-video generation
-      files:
-        - source: "transformer/diffusion_pytorch_model.safetensors"
-          dest: "cogvideox-5b-transformer.safetensors"
-
-    - repo_id: stabilityai/stable-video-diffusion-img2vid
-      description: SVD - 14 frame image-to-video
-      size_gb: 8
-      essential: true
-      category: video
-      type: checkpoints
-      format: fp16
-      vram_gb: 20
-      frames: 14
-      resolution: 576x1024
-      notes: Convert images to short video clips
-      files:
-        - source: "svd.safetensors"
-          dest: "svd.safetensors"
-
-    - repo_id: stabilityai/stable-video-diffusion-img2vid-xt
-      description: SVD-XT - 25 frame image-to-video
-      size_gb: 8
-      essential: false
-      category: video
-      type: checkpoints
-      format: fp16
-      vram_gb: 20
-      frames: 25
-      resolution: 576x1024
-      notes: Extended frame count version
-      files:
-        - source: "svd_xt.safetensors"
-          dest: "svd_xt.safetensors"
-
-  # ==========================================================================
-  # AUDIO GENERATION MODELS
-  # ==========================================================================
-  audio_models:
-    - repo_id: facebook/musicgen-small
-      description: MusicGen Small - Fast generation
-      size_gb: 3
-      essential: false
-      category: audio
-      type: musicgen
-      format: fp32
-      vram_gb: 4
-      duration_seconds: 30
-      notes: Fastest music generation, lower quality
-      files:
-        - source: "pytorch_model.bin"
-          dest: "musicgen-small-pytorch_model.bin"
-
-    - repo_id: facebook/musicgen-medium
-      description: MusicGen Medium - Balanced quality
-      size_gb: 11
-      essential: true
-      category: audio
-      type: musicgen
-      format: fp32
-      vram_gb: 8
-      duration_seconds: 30
-      notes: Best balance of speed and quality
-      files:
-        - source: "pytorch_model.bin"
-          dest: "musicgen-medium-pytorch_model.bin"
-
-    - repo_id: facebook/musicgen-large
-      description: MusicGen Large - Highest quality
-      size_gb: 22
-      essential: false
-      category: audio
-      type: musicgen
-      format: fp32
-      vram_gb: 16
-      duration_seconds: 30
-      notes: Best quality, slower generation
-      files:
-        - source: "pytorch_model-00001-of-00002.bin"
-          dest: "musicgen-large-pytorch_model-00001-of-00002.bin"
-        - source: "pytorch_model-00002-of-00002.bin"
-          dest: "musicgen-large-pytorch_model-00002-of-00002.bin"
-        - source: "pytorch_model.bin.index.json"
-          dest: "musicgen-large-pytorch_model.bin.index.json"
-
-  # ==========================================================================
-  # SUPPORT MODELS (CLIP, IP-Adapter, etc.)
-  # ==========================================================================
-  support_models:
-    - repo_id: openai/clip-vit-large-patch14
-      description: CLIP H - For SD 1.5 IP-Adapter
-      size_gb: 2
-      essential: true
-      category: support
-      type: clip_vision
-      format: fp32
-      vram_gb: 2
-      notes: Text-image understanding model for IP-Adapter
-      files:
-        - source: "model.safetensors"
-          dest: "CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors"
-
-    - repo_id: laion/CLIP-ViT-bigG-14-laion2B-39B-b160k
-      description: CLIP G - For SDXL IP-Adapter
-      size_gb: 7
-      essential: true
-      category: support
-      type: clip_vision
-      format: fp32
-      vram_gb: 4
-      notes: Larger CLIP model for SDXL IP-Adapter
-      files:
-        - source: "open_clip_model.safetensors"
-          dest: "CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors"
-
-    - repo_id: google/siglip-so400m-patch14-384
-      description: SigLIP - For FLUX models
-      size_gb: 2
-      essential: true
-      category: support
-      type: clip_vision
-      format: fp32
-      vram_gb: 2
-      notes: Advanced image-text alignment
-      files:
-        - source: "model.safetensors"
-          dest: "siglip-so400m-patch14-384.safetensors"
-
-  # ==========================================================================
-  # ANIMATEDIFF MODELS
-  # ==========================================================================
-  animatediff_models:
-    - repo_id: guoyww/animatediff
-      description: AnimateDiff Motion Modules
-      size_gb: 2
-      essential: true
-      category: animatediff
-      type: animatediff_models
-      filename: mm_sd_v15
-      format: safetensors
-      vram_gb: 4
-      notes: Motion modules for AnimateDiff text-to-video
-      files:
-        - source: "mm_sd_v15_v2.ckpt"
-          dest: "mm_sd_v15_v2.ckpt"
-
-  # ==========================================================================
-  # CONTROLNET MODELS
-  # ==========================================================================
-  controlnet_models:
-    - repo_id: lllyasviel/control_v11p_sd15_canny
-      description: ControlNet Canny - Edge detection control for SD 1.5
-      size_gb: 1.5
-      essential: false
-      category: controlnet
-      type: controlnet
-      format: safetensors
-      vram_gb: 2
-      notes: Precise edge-based composition control
-      files:
-        - source: "diffusion_pytorch_model.safetensors"
-          dest: "control_v11p_sd15_canny.safetensors"
-
-    - repo_id: lllyasviel/control_v11f1p_sd15_depth
-      description: ControlNet Depth - Depth map control for SD 1.5
-      size_gb: 1.5
-      essential: false
-      category: controlnet
-      type: controlnet
-      format: safetensors
-      vram_gb: 2
-      notes: Depth-based spatial control
-      files:
-        - source: "diffusion_pytorch_model.safetensors"
-          dest: "control_v11p_sd15_depth.safetensors"
-
-    - repo_id: diffusers/controlnet-canny-sdxl-1.0
-      description: ControlNet Canny SDXL - Edge detection for SDXL
-      size_gb: 2.5
-      essential: false
-      category: controlnet
-      type: controlnet
-      format: safetensors
-      vram_gb: 3
-      notes: Canny edge control for SDXL models
-      files:
-        - source: "diffusion_pytorch_model.safetensors"
-          dest: "controlnet-canny-sdxl-1.0.safetensors"
-
-    - repo_id: diffusers/controlnet-depth-sdxl-1.0
-      description: ControlNet Depth SDXL - Depth map for SDXL
-      size_gb: 2.5
-      essential: false
-      category: controlnet
-      type: controlnet
-      format: safetensors
-      vram_gb: 3
-      notes: Depth control for SDXL models
-      files:
-        - source: "diffusion_pytorch_model.safetensors"
-          dest: "controlnet-depth-sdxl-1.0.safetensors"
-
-  # ==========================================================================
-  # IP-ADAPTER MODELS
-  # ==========================================================================
-  ipadapter_models:
-    - repo_id: h94/IP-Adapter
-      description: IP-Adapter SDXL Base - Style & Composition
-      size_gb: 1.3
-      essential: true
-      category: ipadapter
-      type: ipadapter
-      format: safetensors
-      vram_gb: 4
-      notes: Basic IP-Adapter for SDXL
-      files:
-        - source: "sdxl_models/ip-adapter_sdxl.safetensors"
-          dest: "ip-adapter_sdxl.safetensors"
-
-    - repo_id: h94/IP-Adapter
-      description: IP-Adapter SDXL VIT-H - For CLIP-ViT-H
-      size_gb: 0.9
-      essential: true
-      category: ipadapter
-      type: ipadapter
-      format: safetensors
-      vram_gb: 4
-      notes: IP-Adapter for SDXL with VIT-H CLIP vision model
-      files:
-        - source: "sdxl_models/ip-adapter_sdxl_vit-h.safetensors"
-          dest: "ip-adapter_sdxl_vit-h.safetensors"
-
-    - repo_id: h94/IP-Adapter
-      description: IP-Adapter SDXL Plus - High Strength Composition
-      size_gb: 0.9
-      essential: false
-      category: ipadapter
-      type: ipadapter
-      format: safetensors
-      vram_gb: 4
-      notes: Enhanced composition control with higher strength
-      files:
-        - source: "sdxl_models/ip-adapter-plus_sdxl_vit-h.safetensors"
-          dest: "ip-adapter-plus_sdxl_vit-h.safetensors"
-
-    - repo_id: h94/IP-Adapter
-      description: IP-Adapter SDXL Plus Face - Face-focused generation
-      size_gb: 0.5
-      essential: false
-      category: ipadapter
-      type: ipadapter
-      format: safetensors
-      vram_gb: 4
-      notes: Specialized for face transfer and portrait generation
-      files:
-        - source: "sdxl_models/ip-adapter-plus-face_sdxl_vit-h.safetensors"
-          dest: "ip-adapter-plus-face_sdxl_vit-h.safetensors"
-
-# ============================================================================
-# STORAGE & VRAM SUMMARIES
-# ============================================================================
-
-storage_requirements:
-  essential_only:
-    image: 30      # FLUX Schnell + SDXL Base
-    video: 28      # CogVideoX + SVD
-    audio: 11      # MusicGen Medium
-    support: 11    # All 3 CLIP models
-    total: 80      # Total essential storage
-
-  all_models:
-    image: 54      # All image models
-    video: 36      # All video models
-    audio: 36      # All audio models
-    support: 11    # All support models
-    total: 137     # Total with optional models
-
-vram_requirements:
-  # For 24GB GPU (RTX 4090)
-  simultaneous_loadable:
-    - name: Image Focus - FLUX FP16
-      models: [FLUX.1 Schnell]
-      vram_used: 23
-      remaining: 1
-
-    - name: Image Focus - FLUX FP8 + SDXL
-      models: [FLUX.1 Schnell FP8, SDXL Base]
-      vram_used: 24
-      remaining: 0
-
-    - name: Video Generation
-      models: [CogVideoX-5B optimized, SDXL]
-      vram_used: 24
-      remaining: 0
-
-    - name: Multi-Modal
-      models: [SDXL, MusicGen Medium]
-      vram_used: 20
-      remaining: 4
-
-# ============================================================================
-# INSTALLATION PROFILES
-# ============================================================================
-
-installation_profiles:
-  minimal:
-    description: Minimal setup for testing
-    categories: [support_models]
-    storage_gb: 11
-    estimated_time: 5-10 minutes
-
-  essential:
-    description: Essential models only (~80GB)
-    categories: [image_models, video_models, audio_models, support_models]
-    essential_only: true
-    storage_gb: 80
-    estimated_time: 1-2 hours
-
-  image_focused:
-    description: All image generation models
-    categories: [image_models, support_models]
-    storage_gb: 65
-    estimated_time: 45-90 minutes
-
-  video_focused:
-    description: All video generation models
-    categories: [video_models, image_models, support_models]
-    essential_only: true
-    storage_gb: 69
-    estimated_time: 1-2 hours
-
-  complete:
-    description: All models (including optional)
-    categories: [image_models, video_models, audio_models, support_models]
-    storage_gb: 137
-    estimated_time: 2-4 hours
-
-# ============================================================================
-# METADATA
-# ============================================================================
-
-metadata:
-  version: 1.0.0
-  last_updated: 2025-11-21
-  compatible_with:
-    - ComfyUI >= 0.1.0
-    - Python >= 3.10
-    - HuggingFace Hub >= 0.20.0
-  maintainer: Valknar
-  repository: https://github.com/yourusername/runpod

model-orchestrator/models.yaml

@@ -1,77 +0,0 @@
-# Model Registry for AI Orchestrator
-# Add new models by appending to this file
-
-models:
-  # Text Generation Models
-  qwen-2.5-7b:
-    type: text
-    framework: vllm
-    service_script: models/vllm/server.py
-    port: 8000
-    vram_gb: 14
-    startup_time_seconds: 120
-    endpoint: /v1/chat/completions
-    description: "Qwen 2.5 7B Instruct - Fast text generation, no authentication required"
-
-  llama-3.1-8b:
-    type: text
-    framework: vllm
-    service_script: models/vllm/server.py
-    port: 8001
-    vram_gb: 17
-    startup_time_seconds: 120
-    endpoint: /v1/chat/completions
-    description: "Llama 3.1 8B Instruct - Meta's latest model"
-
-# Example: Add more models easily by uncommenting and customizing below
-
-# Future Text Models:
-#  llama-3.1-8b:
-#    type: text
-#    framework: vllm
-#    docker_service: vllm-llama
-#    port: 8004
-#    vram_gb: 17
-#    startup_time_seconds: 120
-#    endpoint: /v1/chat/completions
-#    description: "Llama 3.1 8B Instruct - Meta's latest model"
-
-# Future Image Models:
-#  sdxl:
-#    type: image
-#    framework: openedai-images
-#    docker_service: sdxl
-#    port: 8005
-#    vram_gb: 10
-#    startup_time_seconds: 45
-#    endpoint: /v1/images/generations
-#    description: "Stable Diffusion XL - High quality image generation"
-
-# Future Audio Models:
-#  whisper-large:
-#    type: audio
-#    framework: faster-whisper
-#    docker_service: whisper
-#    port: 8006
-#    vram_gb: 3
-#    startup_time_seconds: 30
-#    endpoint: /v1/audio/transcriptions
-#    description: "Whisper Large v3 - Speech-to-text transcription"
-#
-#  xtts-v2:
-#    type: audio
-#    framework: openedai-speech
-#    docker_service: tts
-#    port: 8007
-#    vram_gb: 3
-#    startup_time_seconds: 30
-#    endpoint: /v1/audio/speech
-#    description: "XTTS v2 - High-quality text-to-speech with voice cloning"
-
-# Configuration
-config:
-  gpu_memory_total_gb: 24
-  allow_concurrent_loading: false  # Sequential loading only
-  model_switch_timeout_seconds: 300  # 5 minutes max for model switching
-  health_check_interval_seconds: 10
-  default_model: qwen-2.5-7b

model-orchestrator/orchestrator.py

@@ -1,404 +0,0 @@
-#!/usr/bin/env python3
-"""
-AI Model Orchestrator for RunPod RTX 4090
-Manages sequential loading of text, image, and music models on a single GPU
-
-Features:
-- Automatic model switching based on request type
-- OpenAI-compatible API endpoints
-- Docker Compose service management
-- GPU memory monitoring
-- Simple YAML configuration for adding new models
-"""
-
-import asyncio
-import logging
-import os
-import time
-from typing import Dict, Optional, Any
-
-import docker
-import httpx
-import yaml
-from fastapi import FastAPI, Request, HTTPException
-from fastapi.responses import JSONResponse, StreamingResponse
-from pydantic import BaseModel
-
-# Configure logging
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-logger = logging.getLogger(__name__)
-
-# FastAPI app
-app = FastAPI(title="AI Model Orchestrator", version="1.0.0")
-
-# Docker client
-docker_client = docker.from_env()
-
-# Global state
-current_model: Optional[str] = None
-model_registry: Dict[str, Dict[str, Any]] = {}
-config: Dict[str, Any] = {}
-
-
-def load_model_registry():
-    """Load model registry from models.yaml"""
-    global model_registry, config
-
-    config_path = os.getenv("MODELS_CONFIG", "/app/models.yaml")
-    logger.info(f"Loading model registry from {config_path}")
-
-    with open(config_path, 'r') as f:
-        data = yaml.safe_load(f)
-
-    model_registry = data.get('models', {})
-    config = data.get('config', {})
-
-    logger.info(f"Loaded {len(model_registry)} models from registry")
-    for model_name, model_info in model_registry.items():
-        logger.info(f"  - {model_name}: {model_info['description']}")
-
-
-def get_docker_service_name(service_name: str) -> str:
-    """Get full Docker service name with project prefix"""
-    project_name = os.getenv("COMPOSE_PROJECT_NAME", "ai")
-    return f"{project_name}_{service_name}_1"
-
-
-async def stop_current_model():
-    """Stop the currently running model service"""
-    global current_model
-
-    if not current_model:
-        logger.info("No model currently running")
-        return
-
-    model_info = model_registry.get(current_model)
-    if not model_info:
-        logger.warning(f"Model {current_model} not found in registry")
-        current_model = None
-        return
-
-    service_name = get_docker_service_name(model_info['docker_service'])
-    logger.info(f"Stopping model: {current_model} (service: {service_name})")
-
-    try:
-        container = docker_client.containers.get(service_name)
-        container.stop(timeout=30)
-        logger.info(f"Stopped {current_model}")
-        current_model = None
-    except docker.errors.NotFound:
-        logger.warning(f"Container {service_name} not found (already stopped?)")
-        current_model = None
-    except Exception as e:
-        logger.error(f"Error stopping {service_name}: {e}")
-        raise
-
-
-async def start_model(model_name: str):
-    """Start a model service"""
-    global current_model
-
-    if model_name not in model_registry:
-        raise HTTPException(status_code=404, detail=f"Model {model_name} not found in registry")
-
-    model_info = model_registry[model_name]
-    service_name = get_docker_service_name(model_info['docker_service'])
-
-    logger.info(f"Starting model: {model_name} (service: {service_name})")
-    logger.info(f"  VRAM requirement: {model_info['vram_gb']} GB")
-    logger.info(f"  Estimated startup time: {model_info['startup_time_seconds']}s")
-
-    try:
-        # Start the container
-        container = docker_client.containers.get(service_name)
-        container.start()
-
-        # Wait for service to be healthy
-        port = model_info['port']
-        endpoint = model_info.get('endpoint', '/')
-        base_url = f"http://localhost:{port}"
-
-        logger.info(f"Waiting for {model_name} to be ready at {base_url}...")
-
-        max_wait = model_info['startup_time_seconds'] + 60  # Add buffer
-        start_time = time.time()
-
-        async with httpx.AsyncClient() as client:
-            while time.time() - start_time < max_wait:
-                try:
-                    # Try health check or root endpoint
-                    health_url = f"{base_url}/health"
-                    try:
-                        response = await client.get(health_url, timeout=5.0)
-                        if response.status_code == 200:
-                            logger.info(f"{model_name} is ready!")
-                            current_model = model_name
-                            return
-                    except:
-                        # Try root endpoint if /health doesn't exist
-                        response = await client.get(base_url, timeout=5.0)
-                        if response.status_code == 200:
-                            logger.info(f"{model_name} is ready!")
-                            current_model = model_name
-                            return
-                except Exception as e:
-                    logger.debug(f"Waiting for {model_name}... ({e})")
-
-                await asyncio.sleep(5)
-
-        raise HTTPException(
-            status_code=503,
-            detail=f"Model {model_name} failed to start within {max_wait}s"
-        )
-
-    except docker.errors.NotFound:
-        raise HTTPException(
-            status_code=500,
-            detail=f"Docker service {service_name} not found. Is it defined in docker-compose?"
-        )
-    except Exception as e:
-        logger.error(f"Error starting {model_name}: {e}")
-        raise HTTPException(status_code=500, detail=str(e))
-
-
-async def ensure_model_running(model_name: str):
-    """Ensure the specified model is running, switching if necessary"""
-    global current_model
-
-    if current_model == model_name:
-        logger.info(f"Model {model_name} already running")
-        return
-
-    logger.info(f"Switching model: {current_model} -> {model_name}")
-
-    # Stop current model
-    await stop_current_model()
-
-    # Start requested model
-    await start_model(model_name)
-
-    logger.info(f"Model switch complete: {model_name} is now active")
-
-
-async def proxy_request(model_name: str, request: Request):
-    """Proxy request to the active model service"""
-    model_info = model_registry[model_name]
-    port = model_info['port']
-
-    # Get request details
-    path = request.url.path
-    method = request.method
-    headers = dict(request.headers)
-    headers.pop('host', None)  # Remove host header
-
-    # Build target URL
-    target_url = f"http://localhost:{port}{path}"
-
-    # Check if this is a streaming request
-    body = await request.body()
-    is_streaming = False
-    if method == "POST" and body:
-        try:
-            import json
-            body_json = json.loads(body)
-            is_streaming = body_json.get('stream', False)
-        except:
-            pass
-
-    logger.info(f"Proxying {method} request to {target_url} (streaming: {is_streaming})")
-
-    if is_streaming:
-        # For streaming requests, use httpx streaming and yield chunks
-        async def stream_response():
-            async with httpx.AsyncClient(timeout=300.0) as client:
-                async with client.stream(method, target_url, content=body, headers=headers) as response:
-                    async for chunk in response.aiter_bytes():
-                        yield chunk
-
-        return StreamingResponse(
-            stream_response(),
-            media_type="text/event-stream",
-            headers={"Cache-Control": "no-cache", "Connection": "keep-alive"}
-        )
-    else:
-        # For non-streaming requests, use the original behavior
-        async with httpx.AsyncClient(timeout=300.0) as client:
-            if method == "GET":
-                response = await client.get(target_url, headers=headers)
-            elif method == "POST":
-                response = await client.post(target_url, content=body, headers=headers)
-            else:
-                raise HTTPException(status_code=405, detail=f"Method {method} not supported")
-
-            # Return response
-            return JSONResponse(
-                content=response.json() if response.headers.get('content-type', '').startswith('application/json') else response.text,
-                status_code=response.status_code,
-                headers=dict(response.headers)
-            )
-
-
-@app.on_event("startup")
-async def startup_event():
-    """Load model registry on startup"""
-    load_model_registry()
-    logger.info("AI Model Orchestrator started successfully")
-    logger.info(f"GPU Memory: {config.get('gpu_memory_total_gb', 24)} GB")
-    logger.info(f"Default model: {config.get('default_model', 'qwen-2.5-7b')}")
-
-
-@app.get("/")
-async def root():
-    """Root endpoint"""
-    return {
-        "service": "AI Model Orchestrator",
-        "version": "1.0.0",
-        "current_model": current_model,
-        "available_models": list(model_registry.keys())
-    }
-
-
-@app.get("/health")
-async def health():
-    """Health check endpoint"""
-    return {
-        "status": "healthy",
-        "current_model": current_model,
-        "model_info": model_registry.get(current_model) if current_model else None,
-        "gpu_memory_total_gb": config.get('gpu_memory_total_gb', 24),
-        "models_available": len(model_registry)
-    }
-
-
-@app.get("/models")
-async def list_models():
-    """List all available models"""
-    return {
-        "models": model_registry,
-        "current_model": current_model
-    }
-
-
-@app.get("/v1/models")
-async def list_models_openai():
-    """OpenAI-compatible models listing endpoint"""
-    models_list = []
-    for model_name, model_info in model_registry.items():
-        models_list.append({
-            "id": model_name,
-            "object": "model",
-            "created": int(time.time()),
-            "owned_by": "pivoine-gpu",
-            "permission": [],
-            "root": model_name,
-            "parent": None,
-        })
-
-    return {
-        "object": "list",
-        "data": models_list
-    }
-
-
-@app.post("/v1/chat/completions")
-async def chat_completions(request: Request):
-    """OpenAI-compatible chat completions endpoint (text models)"""
-    # Parse request to get model name
-    body = await request.json()
-    model_name = body.get('model', config.get('default_model', 'qwen-2.5-7b'))
-
-    # Validate model type
-    if model_name not in model_registry:
-        raise HTTPException(status_code=404, detail=f"Model {model_name} not found")
-
-    if model_registry[model_name]['type'] != 'text':
-        raise HTTPException(status_code=400, detail=f"Model {model_name} is not a text model")
-
-    # Ensure model is running
-    await ensure_model_running(model_name)
-
-    # Proxy request to model
-    return await proxy_request(model_name, request)
-
-
-@app.post("/v1/images/generations")
-async def image_generations(request: Request):
-    """OpenAI-compatible image generation endpoint"""
-    # Parse request to get model name
-    body = await request.json()
-    model_name = body.get('model', 'flux-schnell')
-
-    # Validate model type
-    if model_name not in model_registry:
-        raise HTTPException(status_code=404, detail=f"Model {model_name} not found")
-
-    if model_registry[model_name]['type'] != 'image':
-        raise HTTPException(status_code=400, detail=f"Model {model_name} is not an image model")
-
-    # Ensure model is running
-    await ensure_model_running(model_name)
-
-    # Proxy request to model
-    return await proxy_request(model_name, request)
-
-
-@app.post("/v1/audio/generations")
-async def audio_generations(request: Request):
-    """Custom audio generation endpoint (music/sound effects)"""
-    # Parse request to get model name
-    body = await request.json()
-    model_name = body.get('model', 'musicgen-medium')
-
-    # Validate model type
-    if model_name not in model_registry:
-        raise HTTPException(status_code=404, detail=f"Model {model_name} not found")
-
-    if model_registry[model_name]['type'] != 'audio':
-        raise HTTPException(status_code=400, detail=f"Model {model_name} is not an audio model")
-
-    # Ensure model is running
-    await ensure_model_running(model_name)
-
-    # Proxy request to model
-    return await proxy_request(model_name, request)
-
-
-@app.post("/switch")
-async def switch_model(request: Request):
-    """Manually switch to a specific model"""
-    body = await request.json()
-    model_name = body.get('model')
-
-    if not model_name:
-        raise HTTPException(status_code=400, detail="Model name required")
-
-    if model_name not in model_registry:
-        raise HTTPException(status_code=404, detail=f"Model {model_name} not found")
-
-    await ensure_model_running(model_name)
-
-    return {
-        "status": "success",
-        "model": model_name,
-        "message": f"Switched to {model_name}"
-    }
-
-
-if __name__ == "__main__":
-    import uvicorn
-
-    host = os.getenv("HOST", "0.0.0.0")
-    port = int(os.getenv("PORT", "9000"))
-
-    logger.info(f"Starting AI Model Orchestrator on {host}:{port}")
-
-    uvicorn.run(
-        app,
-        host=host,
-        port=port,
-        log_level="info",
-        access_log=True,
-    )

model-orchestrator/orchestrator_subprocess.py

@@ -1,317 +0,0 @@
-#!/usr/bin/env python3
-"""
-AI Model Orchestrator for RunPod (Process-Based)
-Manages sequential loading of AI models using subprocess instead of Docker
-
-Simplified architecture for RunPod's containerized environment:
-- No Docker-in-Docker complexity
-- Direct process management via subprocess
-- Models run as Python background processes
-- GPU memory efficient (sequential model loading)
-"""
-
-import asyncio
-import logging
-import os
-import subprocess
-import time
-import signal
-from typing import Dict, Optional, Any
-from pathlib import Path
-
-import httpx
-import yaml
-import psutil
-from fastapi import FastAPI, Request, HTTPException
-from fastapi.responses import JSONResponse, StreamingResponse
-from pydantic import BaseModel
-
-# Configure logging
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-logger = logging.getLogger(__name__)
-
-# FastAPI app
-app = FastAPI(title="AI Model Orchestrator (Process-Based)", version="2.0.0")
-
-# Global state
-current_model: Optional[str] = None
-model_processes: Dict[str, subprocess.Popen] = {}
-model_registry: Dict[str, Dict[str, Any]] = {}
-config: Dict[str, Any] = {}
-
-
-def load_model_registry():
-    """Load model registry from models.yaml"""
-    global model_registry, config
-
-    config_path = os.getenv("MODELS_CONFIG", "/workspace/ai/model-orchestrator/models.yaml")
-    logger.info(f"Loading model registry from {config_path}")
-
-    with open(config_path, 'r') as f:
-        data = yaml.safe_load(f)
-
-    model_registry = data.get('models', {})
-    config = data.get('config', {})
-
-    logger.info(f"Loaded {len(model_registry)} models")
-    for model_name, model_config in model_registry.items():
-        logger.info(f"  - {model_name}: {model_config.get('type')} ({model_config.get('framework')})")
-
-
-async def start_model_process(model_name: str) -> bool:
-    """Start a model as a subprocess"""
-    global current_model, model_processes
-
-    if model_name not in model_registry:
-        logger.error(f"Model {model_name} not found in registry")
-        return False
-
-    model_config = model_registry[model_name]
-
-    # Stop current model if running
-    if current_model and current_model != model_name:
-        await stop_model_process(current_model)
-
-    # Check if already running
-    if model_name in model_processes:
-        proc = model_processes[model_name]
-        if proc.poll() is None:  # Still running
-            logger.info(f"Model {model_name} already running")
-            return True
-
-    logger.info(f"Starting model {model_name}...")
-
-    try:
-        # Get service command from config
-        service_script = model_config.get('service_script')
-        if not service_script:
-            logger.error(f"No service_script defined for {model_name}")
-            return False
-
-        script_path = Path(f"/workspace/ai/{service_script}")
-        if not script_path.exists():
-            logger.error(f"Service script not found: {script_path}")
-            return False
-
-        # Start process
-        port = model_config.get('port', 8000)
-        env = os.environ.copy()
-        env.update({
-            'HF_TOKEN': os.getenv('HF_TOKEN', ''),
-            'PORT': str(port),
-            'HOST': '0.0.0.0'
-        })
-
-        proc = subprocess.Popen(
-            ['python3', str(script_path)],
-            env=env,
-            stdout=subprocess.PIPE,
-            stderr=subprocess.PIPE,
-            preexec_fn=os.setsid  # Create new process group
-        )
-
-        model_processes[model_name] = proc
-
-        # Wait for service to be ready
-        max_wait = model_config.get('startup_time_seconds', 120)
-        start_time = time.time()
-
-        while time.time() - start_time < max_wait:
-            if proc.poll() is not None:
-                logger.error(f"Process for {model_name} exited prematurely")
-                return False
-
-            try:
-                async with httpx.AsyncClient() as client:
-                    response = await client.get(
-                        f"http://localhost:{port}/health",
-                        timeout=5.0
-                    )
-                    if response.status_code == 200:
-                        logger.info(f"Model {model_name} is ready on port {port}")
-                        current_model = model_name
-                        return True
-            except:
-                await asyncio.sleep(2)
-
-        logger.error(f"Model {model_name} failed to start within {max_wait}s")
-        await stop_model_process(model_name)
-        return False
-
-    except Exception as e:
-        logger.error(f"Error starting {model_name}: {e}")
-        return False
-
-
-async def stop_model_process(model_name: str):
-    """Stop a running model process"""
-    global model_processes, current_model
-
-    if model_name not in model_processes:
-        logger.warning(f"Model {model_name} not in process registry")
-        return
-
-    proc = model_processes[model_name]
-
-    if proc.poll() is None:  # Still running
-        logger.info(f"Stopping model {model_name}...")
-        try:
-            # Send SIGTERM to process group
-            os.killpg(os.getpgid(proc.pid), signal.SIGTERM)
-
-            # Wait for graceful shutdown
-            try:
-                proc.wait(timeout=10)
-            except subprocess.TimeoutExpired:
-                # Force kill if not terminated
-                os.killpg(os.getpgid(proc.pid), signal.SIGKILL)
-                proc.wait()
-
-            logger.info(f"Model {model_name} stopped")
-        except Exception as e:
-            logger.error(f"Error stopping {model_name}: {e}")
-
-    del model_processes[model_name]
-    if current_model == model_name:
-        current_model = None
-
-
-def get_model_for_endpoint(endpoint: str) -> Optional[str]:
-    """Determine which model handles this endpoint"""
-    for model_name, model_config in model_registry.items():
-        if endpoint.startswith(model_config.get('endpoint', '')):
-            return model_name
-    return None
-
-
-@app.on_event("startup")
-async def startup_event():
-    """Initialize on startup"""
-    logger.info("Starting AI Model Orchestrator (Process-Based)")
-    load_model_registry()
-
-
-@app.on_event("shutdown")
-async def shutdown_event():
-    """Cleanup on shutdown"""
-    logger.info("Shutting down orchestrator...")
-    for model_name in list(model_processes.keys()):
-        await stop_model_process(model_name)
-
-
-@app.get("/health")
-async def health_check():
-    """Health check endpoint"""
-    return {
-        "status": "healthy",
-        "current_model": current_model,
-        "active_processes": len(model_processes),
-        "available_models": list(model_registry.keys())
-    }
-
-
-@app.get("/v1/models")
-async def list_models_openai():
-    """OpenAI-compatible models listing endpoint"""
-    models_list = []
-    for model_name, model_info in model_registry.items():
-        models_list.append({
-            "id": model_name,
-            "object": "model",
-            "created": int(time.time()),
-            "owned_by": "pivoine-gpu",
-            "permission": [],
-            "root": model_name,
-            "parent": None,
-        })
-
-    return {
-        "object": "list",
-        "data": models_list
-    }
-
-
-@app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE", "PATCH"])
-async def proxy_request(request: Request, path: str):
-    """Proxy requests to appropriate model service"""
-    endpoint = f"/{path}"
-
-    # Determine which model should handle this
-    target_model = get_model_for_endpoint(endpoint)
-
-    if not target_model:
-        raise HTTPException(status_code=404, detail=f"No model configured for endpoint: {endpoint}")
-
-    # Ensure model is running
-    if current_model != target_model:
-        logger.info(f"Switching to model {target_model}")
-        success = await start_model_process(target_model)
-        if not success:
-            raise HTTPException(status_code=503, detail=f"Failed to start model {target_model}")
-
-    # Proxy the request
-    model_config = model_registry[target_model]
-    target_url = f"http://localhost:{model_config['port']}/{path}"
-
-    # Get request details
-    method = request.method
-    headers = dict(request.headers)
-    headers.pop('host', None)  # Remove host header
-    body = await request.body()
-
-    # Check if this is a streaming request
-    is_streaming = False
-    if method == "POST" and body:
-        try:
-            import json
-            body_json = json.loads(body)
-            is_streaming = body_json.get('stream', False)
-        except:
-            pass
-
-    logger.info(f"Proxying {method} request to {target_url} (streaming: {is_streaming})")
-
-    try:
-        if is_streaming:
-            # For streaming requests, use httpx streaming and yield chunks
-            async def stream_response():
-                async with httpx.AsyncClient(timeout=300.0) as client:
-                    async with client.stream(method, target_url, content=body, headers=headers) as response:
-                        async for chunk in response.aiter_bytes():
-                            yield chunk
-
-            return StreamingResponse(
-                stream_response(),
-                media_type="text/event-stream",
-                headers={"Cache-Control": "no-cache", "Connection": "keep-alive"}
-            )
-        else:
-            # For non-streaming requests, use the original behavior
-            async with httpx.AsyncClient(timeout=300.0) as client:
-                response = await client.request(
-                    method=method,
-                    url=target_url,
-                    headers=headers,
-                    content=body
-                )
-
-                return JSONResponse(
-                    content=response.json() if response.headers.get('content-type') == 'application/json' else response.text,
-                    status_code=response.status_code
-                )
-    except Exception as e:
-        logger.error(f"Error proxying request: {e}")
-        raise HTTPException(status_code=502, detail=str(e))
-
-
-if __name__ == "__main__":
-    import uvicorn
-
-    port = int(os.getenv("PORT", "9000"))
-    host = os.getenv("HOST", "0.0.0.0")
-
-    logger.info(f"Starting orchestrator on {host}:{port}")
-    uvicorn.run(app, host=host, port=port, log_level="info")

model-orchestrator/requirements.txt

@@ -1,6 +0,0 @@
-fastapi==0.104.1
-uvicorn[standard]==0.24.0
-httpx==0.25.1
-docker==6.1.3
-pyyaml==6.0.1
-pydantic==2.5.0

models_civitai.yaml

@@ -0,0 +1,393 @@
+# ============================================================================
+# CivitAI Model Configuration
+# ============================================================================
+#
+# This configuration file defines all available CivitAI models for download.
+# Models are organized by category: NSFW checkpoints, LoRAs, embeddings, etc.
+#
+# Each model entry contains:
+#   - name: Filesystem-friendly model name
+#   - version_id: CivitAI version ID (used for downloads)
+#   - model_id: CivitAI model ID (used for metadata queries)
+#   - description: Human-readable description
+#   - size_gb: Approximate size in gigabytes
+#   - essential: Whether this is an essential model (true/false)
+#   - type: Model type (checkpoints, loras, embeddings, etc.)
+#   - tags: Array of descriptive tags
+#
+# ============================================================================
+
+# Global settings
+settings:
+  cache_dir: /workspace/models/civitai
+  output_dir: /workspace/ComfyUI/models
+  parallel_downloads: 1
+  retry_attempts: 3
+  timeout_seconds: 3600
+  rate_limit_delay: 5  # seconds between downloads
+
+# Model categories
+model_categories:
+  # ==========================================================================
+  # NSFW IMAGE GENERATION MODELS
+  # ==========================================================================
+  nsfw_checkpoints:
+    - name: lustify-v7-ggwp
+      version_id: "1094291"
+      model_id: "573152"
+      description: "LUSTIFY v7.0 GGWP - Photoreal NSFW checkpoint for women in sexual scenarios"
+      size_gb: 6.31
+      essential: true
+      type: checkpoints
+      format: safetensors
+      base_model: SDXL 1.0
+      vram_gb: 12
+      tags: [nsfw, realistic, photography, women]
+      recommended_settings:
+        sampler: "DPM++ 2M SDE / DPM++ 3M SDE"
+        scheduler: "Exponential / Karras"
+        steps: 30
+        cfg_scale: "4-7"
+      notes: "Understands both danbooru tags and natural language prompting"
+
+    - name: pony-diffusion-v6-xl
+      version_id: "135867"
+      model_id: "257749"
+      description: "Pony Diffusion V6 XL - Versatile anime/cartoon/furry model with balanced content"
+      size_gb: 6.5
+      essential: true
+      type: checkpoints
+      format: safetensors
+      base_model: SDXL 1.0
+      vram_gb: 12
+      tags: [nsfw, anime, furry, cartoon, versatile]
+      files:
+        - source: "ponyDiffusionV6XL_v6StartWithThisOne.safetensors"
+          dest: "ponyDiffusionV6XL_v6StartWithThisOne.safetensors"
+      training_info:
+        images: "2.6M aesthetically ranked"
+        ratio: "1:1:1 safe/questionable/explicit"
+      notes: "Most popular anime/furry model on CivitAI, supports wide range of styles"
+
+    - name: realvisxl-v5
+      version_id: "798204"
+      model_id: "139562"
+      description: "RealVisXL V5.0 Lightning - Photorealistic model for high-quality realistic images"
+      size_gb: 6.8
+      essential: true
+      type: checkpoints
+      format: safetensors
+      base_model: SDXL 1.0
+      vram_gb: 12
+      tags: [nsfw, realistic, photorealistic, professional]
+      recommended_settings:
+        sampler: "DPM++ SDE Karras (4-6 steps for Lightning)"
+      notes: "Excellent for photorealistic portraits and scenes, both SFW and NSFW. Lightning version for faster generation."
+
+    - name: wai-nsfw-illustrious-sdxl
+      version_id: "2167369"
+      model_id: "827184"
+      description: "WAI-NSFW-illustrious-SDXL v15.0 - Actively updated NSFW Illustrious variant"
+      size_gb: 6.5
+      essential: false
+      type: checkpoints
+      format: safetensors
+      base_model: SDXL 1.0
+      vram_gb: 12
+      tags: [nsfw, anime, illustrious, artistic]
+      notes: "Latest version from November 2025, frequently updated"
+
+    - name: talmendoxl
+      version_id: "131869"
+      model_id: "119202"
+      description: "TalmendoXL - Uncensored SDXL model biased towards photorealism"
+      size_gb: 6.6
+      essential: false
+      type: checkpoints
+      format: safetensors
+      base_model: SDXL 1.0
+      vram_gb: 12
+      tags: [nsfw, realistic, uncensored, photorealistic]
+      notes: "Known for high-quality photorealistic NSFW content"
+
+    - name: big-lust-v1-6
+      version_id: "1081768"
+      model_id: "575395"
+      description: "Big Lust v1.6 - SDXL NSFW checkpoint"
+      size_gb: 6.5
+      essential: false
+      type: checkpoints
+      format: safetensors
+      base_model: SDXL 1.0
+      vram_gb: 12
+      tags: [nsfw, versatile]
+      notes: "General-purpose NSFW model with good results across styles"
+
+  # ==========================================================================
+  # NEGATIVE EMBEDDINGS - SD 1.5 (Textual Inversions)
+  # ==========================================================================
+  embeddings_sd15:
+    - name: baddream
+      version_id: "77169"
+      model_id: "72437"
+      description: "BadDream v1.0 - Negative embedding for dreamshaper style artifacts"
+      size_gb: 0.0002
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: SD 1.5
+      tags: [negative, quality, sd15]
+      trigger_word: "BadDream"
+      usage: "embedding:BadDream"
+      notes: "Use in negative prompt. Best combined with UnrealisticDream for optimal results."
+
+    - name: unrealisticdream
+      version_id: "77173"
+      model_id: "72437"
+      description: "UnrealisticDream v1.0 - Improves realistic images, pairs with BadDream"
+      size_gb: 0.0001
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: SD 1.5
+      tags: [negative, realistic, sd15]
+      trigger_word: "UnrealisticDream"
+      usage: "embedding:UnrealisticDream"
+      notes: "Use together with BadDream for best results on realistic images."
+
+    - name: badhandv4
+      version_id: "20068"
+      model_id: "16993"
+      description: "badhandv4 - Improves hand details with minimal style impact"
+      size_gb: 0.00002
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: SD 1.5
+      tags: [negative, hands, anatomy, sd15]
+      trigger_word: "badhandv4"
+      usage: "embedding:badhandv4"
+      recommended_settings:
+        cfg_scale: "≥11"
+      notes: "Designed for AnimeIllustDiffusion but works with other SD 1.5 models. Better results at higher CFG."
+
+    - name: easynegative
+      version_id: "9208"
+      model_id: "7808"
+      description: "EasyNegative - General-purpose negative embedding"
+      size_gb: 0.00002
+      essential: false
+      type: embeddings
+      format: safetensors
+      base_model: SD 1.5
+      tags: [negative, quality, sd15]
+      trigger_word: "easynegative"
+      usage: "embedding:easynegative"
+      notes: "Optimized for Counterfeit model. Effectiveness varies by model."
+
+    - name: fastnegativev2
+      version_id: "94057"
+      model_id: "71961"
+      description: "FastNegativeV2 - Token mix of common negative prompts"
+      size_gb: 0.0002
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: SD 1.5
+      tags: [negative, quality, sd15]
+      trigger_word: "FastNegativeV2"
+      usage: "embedding:FastNegativeV2"
+      notes: "V2 is balanced and doesn't require weight reduction. Use at strength 1.0."
+
+    - name: badneganatomy
+      version_id: "64063"
+      model_id: "59614"
+      description: "BadNegAnatomyV1-neg - Improves aesthetics and character anatomy"
+      size_gb: 0.0002
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: SD 1.5
+      tags: [negative, anatomy, quality, sd15]
+      trigger_word: "BadNegAnatomyV1-neg"
+      usage: "embedding:BadNegAnatomyV1-neg"
+      notes: "Particularly improves body joints and overall anatomy."
+
+  # ==========================================================================
+  # NEGATIVE EMBEDDINGS - SDXL (Textual Inversions)
+  # ==========================================================================
+  embeddings_sdxl:
+    - name: badx-sdxl
+      version_id: "981304"
+      model_id: "122403"
+      description: "Bad X v1.1 - SDXL negative embedding for anatomy and realism"
+      size_gb: 0.000004
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: SDXL 1.0
+      tags: [negative, quality, anatomy, sdxl]
+      trigger_word: "BadX"
+      usage: "embedding:BadX"
+      files:
+        - source: "BadX-neg.pt"
+          dest: "BadX-neg.pt"
+      notes: "Use in negative prompt with LUSTIFY, RealVisXL, or other SDXL checkpoints. Fixes facial/hand artifacts."
+
+  # ==========================================================================
+  # NEGATIVE EMBEDDINGS - Pony SDXL (Textual Inversions)
+  # ==========================================================================
+  embeddings_pony:
+    - name: pony-pdxl-hq-v3
+      version_id: "720175"
+      model_id: "332646"
+      description: "Pony PDXL High Quality V3 (zPDXL3) - Quality enhancer for Pony models"
+      size_gb: 0.0003
+      essential: true
+      type: embeddings
+      format: safetensors
+      base_model: Pony SDXL
+      tags: [negative, quality, pony, nsfw]
+      trigger_word: "zPDXL3"
+      usage: "embedding:zPDXL3"
+      files:
+        - source: "zPDXL3.safetensors"
+          dest: "zPDXL3.safetensors"
+      recommended_settings:
+        strength: "1.0-2.0"
+      notes: "ONLY works with Pony Diffusion models. Removes censoring and improves quality."
+
+    - name: pony-pdxl-xxx
+      version_id: "380277"
+      model_id: "332646"
+      description: "Pony PDXL XXX Rating (zPDXLxxx) - Enables NSFW content"
+      size_gb: 0.00003
+      essential: true
+      type: embeddings
+      format: pt
+      base_model: Pony SDXL
+      tags: [negative, nsfw, pony]
+      trigger_word: "zPDXLxxx"
+      usage: "embedding:zPDXLxxx"
+      files:
+        - source: "zPDXLxxx.pt"
+          dest: "zPDXLxxx.pt"
+      recommended_settings:
+        strength: "1.0-2.0"
+      notes: "ONLY for Pony Diffusion models. Enables explicit NSFW content generation."
+
+  # ==========================================================================
+  # LORA MODELS (Style Enhancement & Detail)
+  # ==========================================================================
+  loras:
+    - name: add-detail-xl
+      version_id: "135867"
+      model_id: "122359"
+      description: "Add Detail XL - Detail enhancement LoRA for SDXL"
+      size_gb: 0.21
+      essential: true
+      type: loras
+      format: safetensors
+      base_model: SDXL 1.0
+      tags: [enhancement, detail, quality]
+      recommended_settings:
+        strength_model: "0.5-1.0"
+        strength_clip: "0.5-1.0"
+      notes: "Popular detail enhancer LoRA, works with all SDXL checkpoints"
+
+    - name: siesta-v1-1
+      version_id: "unknown"
+      model_id: "unknown"
+      description: "Siesta v1.1 - Style LoRA"
+      size_gb: 0.04
+      essential: false
+      type: loras
+      format: safetensors
+      base_model: SDXL 1.0
+      tags: [style, artistic]
+      notes: "Manually downloaded LoRA, version_id needs to be updated"
+
+# ============================================================================
+# STORAGE & VRAM SUMMARIES
+# ============================================================================
+
+storage_requirements:
+  essential_only:
+    nsfw_checkpoints: 19.61  # LUSTIFY + Pony + RealVisXL
+    total: 19.61
+
+  all_models:
+    nsfw_checkpoints: 39.21  # All 6 checkpoints
+    total: 39.21
+
+vram_requirements:
+  # For 24GB GPU (RTX 4090)
+  notes: |
+    All SDXL checkpoints require approximately 12GB VRAM.
+    Can run one checkpoint at a time on 24GB GPU.
+    For simultaneous loading with other models (ComfyUI, vLLM),
+    ensure combined VRAM usage stays under 24GB.
+
+# ============================================================================
+# INSTALLATION PROFILES
+# ============================================================================
+
+installation_profiles:
+  essential:
+    description: Essential NSFW models only (3 checkpoints)
+    categories: [nsfw_checkpoints]
+    essential_only: true
+    storage_gb: 19.61
+    estimated_time: "30-60 minutes"
+    models:
+      - lustify-v7-ggwp
+      - pony-diffusion-v6-xl
+      - realvisxl-v5
+
+  complete:
+    description: All NSFW models (6 checkpoints)
+    categories: [nsfw_checkpoints]
+    storage_gb: 39.21
+    estimated_time: "1-2 hours"
+    models:
+      - lustify-v7-ggwp
+      - pony-diffusion-v6-xl
+      - realvisxl-v5
+      - wai-nsfw-illustrious-sdxl
+      - talmendoxl
+      - big-lust-v1-6
+
+  realistic_only:
+    description: Photorealistic NSFW models only
+    categories: [nsfw_checkpoints]
+    storage_gb: 13.11
+    estimated_time: "20-40 minutes"
+    models:
+      - lustify-v7-ggwp
+      - realvisxl-v5
+
+  anime_only:
+    description: Anime/cartoon NSFW models only
+    categories: [nsfw_checkpoints]
+    storage_gb: 13.0
+    estimated_time: "20-40 minutes"
+    models:
+      - pony-diffusion-v6-xl
+      - wai-nsfw-illustrious-sdxl
+
+# ============================================================================
+# METADATA
+# ============================================================================
+
+metadata:
+  version: 1.0.0
+  last_updated: 2025-11-23
+  compatible_with:
+    - ComfyUI >= 0.1.0
+    - SDXL base models
+  maintainer: Valknar
+  repository: https://github.com/valknar/runpod
+  notes: |
+    All models listed are NSFW (adult content) and hosted on CivitAI.
+    Requires CIVITAI_API_KEY for downloads.
+    Models will be downloaded to cache_dir and symlinked to ComfyUI.

models_huggingface.yaml

@@ -0,0 +1,715 @@
+settings:
+  cache_dir: /workspace/huggingface_cache
+  parallel_downloads: 1
+  retry_attempts: 3
+  timeout_seconds: 3600
+model_categories:
+  image_models:
+  - repo_id: black-forest-labs/FLUX.1-schnell
+    description: FLUX.1 Schnell - Fast 4-step inference
+    size_gb: 23
+    essential: true
+    category: image
+    format: fp16
+    vram_gb: 23
+    notes: Industry-leading image generation quality
+    files:
+    - source: flux1-schnell.safetensors
+      dest: unet/flux1-schnell.safetensors
+  - repo_id: black-forest-labs/FLUX.1-dev
+    description: FLUX.1 Dev - Balanced quality/speed
+    size_gb: 23
+    essential: false
+    category: image
+    format: fp16
+    vram_gb: 23
+    notes: Development version with enhanced features
+    files:
+    - source: flux1-dev.safetensors
+      dest: unet/flux1-dev.safetensors
+  - repo_id: runwayml/stable-diffusion-v1-5
+    description: SD 1.5 - For AnimateDiff
+    size_gb: 4
+    essential: true
+    category: image
+    format: fp16
+    vram_gb: 8
+    notes: Stable Diffusion 1.5 required for AnimateDiff motion modules
+    files:
+    - source: v1-5-pruned-emaonly.safetensors
+      dest: checkpoints/v1-5-pruned-emaonly.safetensors
+  - repo_id: stabilityai/stable-diffusion-xl-base-1.0
+    description: SDXL Base 1.0 - Industry standard
+    size_gb: 7
+    essential: true
+    category: image
+    format: fp16
+    vram_gb: 12
+    notes: Most widely used Stable Diffusion model
+    files:
+    - source: sd_xl_base_1.0.safetensors
+      dest: checkpoints/sd_xl_base_1.0.safetensors
+  - repo_id: stabilityai/stable-diffusion-xl-refiner-1.0
+    description: SDXL Refiner 1.0 - Enhances base output
+    size_gb: 6
+    essential: false
+    category: image
+    format: fp16
+    vram_gb: 12
+    notes: Use after SDXL base for improved details
+    files:
+    - source: sd_xl_refiner_1.0.safetensors
+      dest: checkpoints/sd_xl_refiner_1.0.safetensors
+  - repo_id: stabilityai/stable-diffusion-3.5-large
+    description: SD 3.5 Large Complete - Checkpoint and text encoders
+    size_gb: 31
+    essential: false
+    category: image
+    format: mixed
+    vram_gb: 20
+    notes: Complete SD3.5 Large model with checkpoint and all text encoders (CLIP-L,
+      CLIP-G, T5-XXL)
+    files:
+    - source: sd3.5_large.safetensors
+      dest: checkpoints/sd3.5_large.safetensors
+    - source: text_encoders/clip_l.safetensors
+      dest: checkpoints/clip_l.safetensors
+    - source: text_encoders/clip_g.safetensors
+      dest: checkpoints/clip_g.safetensors
+    - source: text_encoders/t5xxl_fp16.safetensors
+      dest: checkpoints/t5xxl_fp16.safetensors
+  - repo_id: John6666/diving-illustrious-real-asian-v50-sdxl
+    description: Diving Illustrious Real Asian v5.0 - Photorealistic Asian subjects
+    size_gb: 7
+    essential: false
+    category: image
+    format: fp16
+    vram_gb: 12
+    notes: SDXL fine-tune specializing in photorealistic Asian subjects with illustrious
+      quality
+    files:
+    - source: unet/diffusion_pytorch_model.safetensors
+      dest: checkpoints/diving-illustrious-real-asian-v50-sdxl.safetensors
+  - repo_id: playgroundai/playground-v2.5-1024px-aesthetic
+    description: Playground v2.5 - 1024px aesthetic images
+    size_gb: 7
+    essential: false
+    category: image
+    format: fp16
+    vram_gb: 12
+    notes: Highly aesthetic 1024x1024 outputs, outperforms SDXL and DALL-E 3 in user
+      studies
+    files:
+    - source: playground-v2.5-1024px-aesthetic.fp16.safetensors
+      dest: checkpoints/playground-v2.5-1024px-aesthetic.safetensors
+  - repo_id: Lykon/dreamshaper-8
+    description: DreamShaper 8 - Multi-style versatile model
+    size_gb: 4
+    essential: false
+    category: image
+    format: fp16
+    vram_gb: 8
+    notes: Versatile SD1.5 fine-tune balancing photorealistic and anime styles with
+      strong LoRA support
+    files:
+    - source: unet/diffusion_pytorch_model.fp16.safetensors
+      dest: checkpoints/dreamshaper-8.safetensors
+  video_models:
+  - repo_id: THUDM/CogVideoX-5b
+    description: CogVideoX-5B - Professional text-to-video
+    size_gb: 20
+    essential: true
+    category: video
+    format: fp16
+    vram_gb: 20
+    frames: 49
+    resolution: 720p
+    notes: State-of-the-art text-to-video generation, auto-downloaded by DownloadAndLoadCogVideoModel
+      node
+    files:
+    - source: transformer/diffusion_pytorch_model-00001-of-00002.safetensors
+      dest: diffusion_models/cogvideox-5b-transformer-00001-of-00002.safetensors
+    - source: transformer/diffusion_pytorch_model-00002-of-00002.safetensors
+      dest: diffusion_models/cogvideox-5b-transformer-00002-of-00002.safetensors
+    - source: transformer/diffusion_pytorch_model.safetensors.index.json
+      dest: diffusion_models/cogvideox-5b-transformer.safetensors.index.json
+  - repo_id: THUDM/CogVideoX-5b-I2V
+    description: CogVideoX-5B-I2V - Image-to-video generation
+    size_gb: 20
+    essential: true
+    category: video
+    format: fp16
+    vram_gb: 20
+    frames: 49
+    resolution: 720p
+    notes: Image-to-video model, auto-downloaded by DownloadAndLoadCogVideoModel node
+    files:
+    - source: transformer/diffusion_pytorch_model-00001-of-00003.safetensors
+      dest: diffusion_models/cogvideox-5b-i2v-transformer-00001-of-00003.safetensors
+    - source: transformer/diffusion_pytorch_model-00002-of-00003.safetensors
+      dest: diffusion_models/cogvideox-5b-i2v-transformer-00002-of-00003.safetensors
+    - source: transformer/diffusion_pytorch_model-00003-of-00003.safetensors
+      dest: diffusion_models/cogvideox-5b-i2v-transformer-00003-of-00003.safetensors
+    - source: transformer/diffusion_pytorch_model.safetensors.index.json
+      dest: diffusion_models/cogvideox-5b-i2v-transformer.safetensors.index.json
+  - repo_id: stabilityai/stable-video-diffusion-img2vid
+    description: SVD - 14 frame image-to-video
+    size_gb: 8
+    essential: true
+    category: video
+    format: fp16
+    vram_gb: 20
+    frames: 14
+    resolution: 576x1024
+    notes: Convert images to short video clips
+    files:
+    - source: svd.safetensors
+      dest: checkpoints/svd.safetensors
+  - repo_id: stabilityai/stable-video-diffusion-img2vid-xt
+    description: SVD-XT - 25 frame image-to-video
+    size_gb: 8
+    essential: false
+    category: video
+    format: fp16
+    vram_gb: 20
+    frames: 25
+    resolution: 576x1024
+    notes: Extended frame count version
+    files:
+    - source: svd_xt.safetensors
+      dest: checkpoints/svd_xt.safetensors
+  - repo_id: Comfy-Org/HunyuanVideo_repackaged
+    description: HunyuanVideo Complete - 720p T2V/I2V models with VAE and encoders
+    size_gb: 51
+    essential: true
+    category: video
+    format: bf16
+    vram_gb: 24
+    frames: 129
+    resolution: 720p
+    notes: Complete HunyuanVideo family - T2V, I2V v1/v2, 3D VAE, LLaVA LLaMA3 text/vision
+      encoders
+    files:
+    - source: split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors
+      dest: diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors
+    - source: split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors
+      dest: diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors
+    - source: split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors
+      dest: diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors
+    - source: split_files/vae/hunyuan_video_vae_bf16.safetensors
+      dest: diffusion_models/hunyuan_video_vae_bf16.safetensors
+    - source: split_files/text_encoders/llava_llama3_fp8_scaled.safetensors
+      dest: diffusion_models/llava_llama3_fp8_scaled.safetensors
+    - source: split_files/clip_vision/llava_llama3_vision.safetensors
+      dest: diffusion_models/llava_llama3_vision.safetensors
+  - repo_id: Comfy-Org/HunyuanVideo_1.5_repackaged
+    description: HunyuanVideo 1.5 Complete - 720p/1080p T2V/SR with encoders
+    size_gb: 51.5
+    essential: true
+    category: video
+    format: fp16
+    vram_gb: 24
+    frames: 129-257
+    resolution: 720p-1080p
+    notes: Complete HunyuanVideo 1.5 - T2V 720p, SR 1080p, VAE, Qwen 2.5 VL, ByT5
+      GlyphXL encoders
+    files:
+    - source: hunyuanvideo1.5_720p_t2v_fp16.safetensors
+      dest: diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors
+    - source: hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors
+      dest: diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors
+    - source: hunyuanvideo15_vae_fp16.safetensors
+      dest: diffusion_models/hunyuanvideo15_vae_fp16.safetensors
+    - source: qwen_2.5_vl_7b_fp8_scaled.safetensors
+      dest: diffusion_models/qwen_2.5_vl_7b_fp8_scaled.safetensors
+    - source: byt5_small_glyphxl_fp16.safetensors
+      dest: diffusion_models/byt5_small_glyphxl_fp16.safetensors
+  - repo_id: Comfy-Org/Wan_2.2_ComfyUI_Repackaged
+    description: Wan2.2 Complete - All video models, VAEs, and LoRAs
+    size_gb: 220
+    essential: true
+    category: video
+    format: mixed
+    vram_gb: 24
+    frames: 81
+    resolution: 640x640
+    notes: Complete Wan2.2 model family - TI2V 5B, T2V 14B, I2V 14B, Animate, S2V,
+      Fun Inpaint/Control/Camera, VAEs, CLIP Vision H, Wav2Vec2, and LoRA accelerators
+    files:
+    - source: wan2.2_ti2v_5B_fp16.safetensors
+      dest: diffusion_models/wan2.2_ti2v_5B_fp16.safetensors
+    - source: wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_i2v_high_noise_14B_fp16.safetensors
+      dest: diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors
+    - source: wan2.2_i2v_low_noise_14B_fp16.safetensors
+      dest: diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors
+    - source: wan2.2_animate_14B_bf16.safetensors
+      dest: diffusion_models/wan2.2_animate_14B_bf16.safetensors
+    - source: wan2.2_s2v_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors
+    - source: wan2.2_s2v_14B_bf16.safetensors
+      dest: diffusion_models/wan2.2_s2v_14B_bf16.safetensors
+    - source: wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors
+      dest: diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors
+    - source: wan2.2_vae.safetensors
+      dest: diffusion_models/wan2.2_vae.safetensors
+    - source: wan_2.1_vae.safetensors
+      dest: diffusion_models/wan_2.1_vae.safetensors
+    - source: clip_vision_h.safetensors
+      dest: diffusion_models/clip_vision_h.safetensors
+    - source: wav2vec2_large_english_fp16.safetensors
+      dest: diffusion_models/wav2vec2_large_english_fp16.safetensors
+    - source: lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
+      dest: diffusion_models/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
+    - source: wan2.2_t2v_lightx2v_4steps_lora_v1.1_high_noise.safetensors
+      dest: diffusion_models/wan2.2_t2v_lightx2v_4steps_lora_v1.1_high_noise.safetensors
+    - source: wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors
+      dest: diffusion_models/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors
+    - source: wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors
+      dest: diffusion_models/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors
+  audio_models:
+  - repo_id: facebook/musicgen-small
+    description: MusicGen Small - Fast generation
+    size_gb: 3
+    essential: false
+    category: audio
+    format: fp32
+    vram_gb: 4
+    duration_seconds: 30
+    notes: Fastest music generation, lower quality
+    files:
+    - source: pytorch_model.bin
+      dest: musicgen/musicgen-small-pytorch_model.bin
+  - repo_id: facebook/musicgen-medium
+    description: MusicGen Medium - Balanced quality
+    size_gb: 11
+    essential: true
+    category: audio
+    format: fp32
+    vram_gb: 8
+    duration_seconds: 30
+    notes: Best balance of speed and quality
+    files:
+    - source: pytorch_model.bin
+      dest: musicgen/musicgen-medium-pytorch_model.bin
+  - repo_id: facebook/musicgen-large
+    description: MusicGen Large - Highest quality
+    size_gb: 22
+    essential: false
+    category: audio
+    format: fp32
+    vram_gb: 16
+    duration_seconds: 30
+    notes: Best quality, slower generation
+    files:
+    - source: pytorch_model-00001-of-00002.bin
+      dest: musicgen/musicgen-large-pytorch_model-00001-of-00002.bin
+    - source: pytorch_model-00002-of-00002.bin
+      dest: musicgen/musicgen-large-pytorch_model-00002-of-00002.bin
+    - source: pytorch_model.bin.index.json
+      dest: musicgen/musicgen-large-pytorch_model.bin.index.json
+  - repo_id: Comfy-Org/ACE-Step_ComfyUI_repackaged
+    description: ACE Step v1 3.5B - Fast coherent music generation with 19-language
+      support
+    size_gb: 7.7
+    essential: true
+    category: audio
+    format: safetensors
+    vram_gb: 16
+    duration_seconds: 240
+    notes: 15x faster than LLM baselines, superior structural coherence, voice cloning,
+      19-language lyrics
+    files:
+    - source: all_in_one/ace_step_v1_3.5b.safetensors
+      dest: checkpoints/ace_step_v1_3.5b.safetensors
+  - repo_id: ACE-Step/ACE-Step-v1-chinese-rap-LoRA
+    description: ACE Step Chinese RAP LoRA - Enhanced Chinese pronunciation and hip-hop
+      genre
+    size_gb: 0.3
+    essential: false
+    category: audio
+    format: safetensors
+    notes: Improves Chinese pronunciation accuracy and hip-hop/electronic genre adherence
+    files:
+    - source: pytorch_lora_weights.safetensors
+      dest: loras/ace-step-chinese-rap-lora.safetensors
+  support_models:
+  - repo_id: openai/clip-vit-large-patch14
+    description: CLIP H - For SD 1.5 IP-Adapter
+    size_gb: 2
+    essential: true
+    category: support
+    format: fp32
+    vram_gb: 2
+    notes: Text-image understanding model for IP-Adapter
+    files:
+    - source: model.safetensors
+      dest: clip_vision/CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors
+  - repo_id: laion/CLIP-ViT-bigG-14-laion2B-39B-b160k
+    description: CLIP G - For SDXL IP-Adapter
+    size_gb: 7
+    essential: true
+    category: support
+    format: fp32
+    vram_gb: 4
+    notes: Larger CLIP model for SDXL IP-Adapter
+    files:
+    - source: open_clip_model.safetensors
+      dest: clip_vision/CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors
+  - repo_id: google/siglip-so400m-patch14-384
+    description: SigLIP - For FLUX models
+    size_gb: 2
+    essential: true
+    category: support
+    format: fp32
+    vram_gb: 2
+    notes: Advanced image-text alignment
+    files:
+    - source: model.safetensors
+      dest: clip_vision/siglip-so400m-patch14-384.safetensors
+  - repo_id: black-forest-labs/FLUX.1-schnell
+    description: FLUX VAE - Autoencoder for FLUX models
+    size_gb: 0.5
+    essential: true
+    category: support
+    format: safetensors
+    vram_gb: 1
+    notes: VAE autoencoder required for FLUX image decoding
+    files:
+    - source: ae.safetensors
+      dest: vae/ae.safetensors
+  - repo_id: ai-forever/Real-ESRGAN
+    description: RealESRGAN x2 - 2x upscaling model
+    size_gb: 0.06
+    essential: true
+    category: support
+    format: pth
+    vram_gb: 2
+    notes: Fast 2x upscaling model for general purpose enhancement
+    files:
+    - source: RealESRGAN_x2.pth
+      dest: upscale_models/RealESRGAN_x2.pth
+  - repo_id: ai-forever/Real-ESRGAN
+    description: RealESRGAN x4 - 4x upscaling model
+    size_gb: 0.06
+    essential: true
+    category: support
+    format: pth
+    vram_gb: 4
+    notes: High-quality 4x upscaling model for detail enhancement
+    files:
+    - source: RealESRGAN_x4.pth
+      dest: upscale_models/RealESRGAN_x4.pth
+  - repo_id: Comfy-Org/Wan_2.1_ComfyUI_repackaged
+    description: UMT5-XXL FP8 - Text encoder for all Wan2.2 models
+    size_gb: 10
+    essential: true
+    category: support
+    format: fp8_scaled
+    vram_gb: 5
+    notes: Shared text encoder for all Wan2.2 models (5B and 14B), FP8 quantized
+    files:
+    - source: umt5_xxl_fp8_e4m3fn_scaled.safetensors
+      dest: text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
+  animatediff_models:
+  - repo_id: guoyww/animatediff
+    description: AnimateDiff Motion Modules
+    size_gb: 2
+    essential: true
+    category: animatediff
+    filename: mm_sd_v15
+    format: safetensors
+    vram_gb: 4
+    notes: Motion modules for AnimateDiff text-to-video
+    files:
+    - source: mm_sd_v15_v2.ckpt
+      dest: animatediff_models/mm_sd_v15_v2.ckpt
+  controlnet_models:
+  - repo_id: lllyasviel/control_v11p_sd15_canny
+    description: ControlNet Canny - Edge detection control for SD 1.5
+    size_gb: 1.5
+    essential: false
+    category: controlnet
+    format: safetensors
+    vram_gb: 2
+    notes: Precise edge-based composition control
+    files:
+    - source: diffusion_pytorch_model.safetensors
+      dest: controlnet/control_v11p_sd15_canny.safetensors
+  - repo_id: lllyasviel/control_v11f1p_sd15_depth
+    description: ControlNet Depth - Depth map control for SD 1.5
+    size_gb: 1.5
+    essential: false
+    category: controlnet
+    format: safetensors
+    vram_gb: 2
+    notes: Depth-based spatial control
+    files:
+    - source: diffusion_pytorch_model.safetensors
+      dest: controlnet/control_v11p_sd15_depth.safetensors
+  - repo_id: diffusers/controlnet-canny-sdxl-1.0
+    description: ControlNet Canny SDXL - Edge detection for SDXL
+    size_gb: 2.5
+    essential: false
+    category: controlnet
+    format: safetensors
+    vram_gb: 3
+    notes: Canny edge control for SDXL models
+    files:
+    - source: diffusion_pytorch_model.safetensors
+      dest: controlnet/controlnet-canny-sdxl-1.0.safetensors
+  - repo_id: diffusers/controlnet-depth-sdxl-1.0
+    description: ControlNet Depth SDXL - Depth map for SDXL
+    size_gb: 2.5
+    essential: false
+    category: controlnet
+    format: safetensors
+    vram_gb: 3
+    notes: Depth control for SDXL models
+    files:
+    - source: diffusion_pytorch_model.safetensors
+      dest: controlnet/controlnet-depth-sdxl-1.0.safetensors
+  ipadapter_models:
+  - repo_id: h94/IP-Adapter
+    description: IP-Adapter SDXL Base - Style & Composition
+    size_gb: 1.3
+    essential: true
+    category: ipadapter
+    format: safetensors
+    vram_gb: 4
+    notes: Basic IP-Adapter for SDXL
+    files:
+    - source: sdxl_models/ip-adapter_sdxl.safetensors
+      dest: ipadapter/ip-adapter_sdxl.safetensors
+  - repo_id: h94/IP-Adapter
+    description: IP-Adapter SDXL VIT-H - For CLIP-ViT-H
+    size_gb: 0.9
+    essential: true
+    category: ipadapter
+    format: safetensors
+    vram_gb: 4
+    notes: IP-Adapter for SDXL with VIT-H CLIP vision model
+    files:
+    - source: sdxl_models/ip-adapter_sdxl_vit-h.safetensors
+      dest: ipadapter/ip-adapter_sdxl_vit-h.safetensors
+  - repo_id: h94/IP-Adapter
+    description: IP-Adapter SDXL Plus - High Strength Composition
+    size_gb: 0.9
+    essential: false
+    category: ipadapter
+    format: safetensors
+    vram_gb: 4
+    notes: Enhanced composition control with higher strength
+    files:
+    - source: sdxl_models/ip-adapter-plus_sdxl_vit-h.safetensors
+      dest: ipadapter/ip-adapter-plus_sdxl_vit-h.safetensors
+  - repo_id: h94/IP-Adapter
+    description: IP-Adapter SDXL Plus Face - Face-focused generation
+    size_gb: 0.5
+    essential: false
+    category: ipadapter
+    format: safetensors
+    vram_gb: 4
+    notes: Specialized for face transfer and portrait generation
+    files:
+    - source: sdxl_models/ip-adapter-plus-face_sdxl_vit-h.safetensors
+      dest: ipadapter/ip-adapter-plus-face_sdxl_vit-h.safetensors
+  diffrhythm_models:
+  - repo_id: ASLP-lab/DiffRhythm-1_2
+    description: DiffRhythm 1.2 - 95 second generation model
+    size_gb: 2
+    essential: true
+    category: diffrhythm
+    format: pt
+    vram_gb: 12
+    duration_seconds: 95
+    notes: Latest 95-second generation model
+    files:
+    - source: cfm_model.pt
+      dest: TTS/DiffRhythm/cfm_model_v1_2.pt
+  - repo_id: ASLP-lab/DiffRhythm-full
+    description: DiffRhythm Full - 4m45s full-length generation
+    size_gb: 2
+    essential: false
+    category: diffrhythm
+    format: pt
+    vram_gb: 16
+    duration_seconds: 285
+    notes: Full-length 4 minute 45 second music generation
+    files:
+    - source: cfm_model.pt
+      dest: TTS/DiffRhythm/cfm_full_model.pt
+  - repo_id: ASLP-lab/DiffRhythm-base
+    description: DiffRhythm Base - 95 second base model
+    size_gb: 2
+    essential: false
+    category: diffrhythm
+    format: pt
+    vram_gb: 12
+    duration_seconds: 95
+    notes: Base 95-second model
+    files:
+    - source: cfm_model.pt
+      dest: TTS/DiffRhythm/cfm_model.pt
+  - repo_id: ASLP-lab/DiffRhythm-vae
+    description: DiffRhythm VAE - Variational autoencoder
+    size_gb: 1
+    essential: true
+    category: diffrhythm
+    format: pt
+    vram_gb: 2
+    notes: VAE component fine-tuned from Stable Audio Open (Stability AI Community
+      License)
+    files:
+    - source: vae_model.pt
+      dest: TTS/DiffRhythm/vae_model.pt
+  - repo_id: OpenMuQ/MuQ-MuLan-large
+    description: MuQ-MuLan-large - Music-text joint embedding (~700M parameters)
+    size_gb: 3
+    essential: true
+    category: diffrhythm
+    format: bin
+    vram_gb: 4
+    notes: Music-text joint embedding for semantic understanding (English/Chinese)
+    files:
+    - source: config.json
+      dest: TTS/DiffRhythm/MuQ-MuLan-large/config.json
+    - source: pytorch_model.bin
+      dest: TTS/DiffRhythm/MuQ-MuLan-large/pytorch_model.bin
+  - repo_id: OpenMuQ/MuQ-large-msd-iter
+    description: MuQ-large-msd-iter - Music representation learning (~300M parameters)
+    size_gb: 1.2
+    essential: true
+    category: diffrhythm
+    format: safetensors
+    vram_gb: 2
+    notes: Music representation model trained on Million Song Dataset
+    files:
+    - source: config.json
+      dest: TTS/DiffRhythm/MuQ-large-msd-iter/config.json
+    - source: model.safetensors
+      dest: TTS/DiffRhythm/MuQ-large-msd-iter/model.safetensors
+  - repo_id: FacebookAI/xlm-roberta-base
+    description: XLM-RoBERTa Base - Multilingual text encoder (100 languages, 0.3B
+      params)
+    size_gb: 1.1
+    essential: true
+    category: diffrhythm
+    format: safetensors
+    vram_gb: 1
+    notes: Multilingual text encoding for 100 languages
+    files:
+    - source: config.json
+      dest: TTS/DiffRhythm/xlm-roberta-base/config.json
+    - source: model.safetensors
+      dest: TTS/DiffRhythm/xlm-roberta-base/model.safetensors
+    - source: sentencepiece.bpe.model
+      dest: TTS/DiffRhythm/xlm-roberta-base/sentencepiece.bpe.model
+    - source: tokenizer.json
+      dest: TTS/DiffRhythm/xlm-roberta-base/tokenizer.json
+    - source: tokenizer_config.json
+      dest: TTS/DiffRhythm/xlm-roberta-base/tokenizer_config.json
+storage_requirements:
+  essential_only:
+    image: 30
+    video: 28
+    audio: 11
+    support: 11
+    diffrhythm: 10
+    total: 90
+  all_models:
+    image: 54
+    video: 36
+    audio: 36
+    support: 11
+    diffrhythm: 12
+    total: 149
+vram_requirements:
+  simultaneous_loadable:
+  - name: Image Focus - FLUX FP16
+    models:
+    - FLUX.1 Schnell
+    vram_used: 23
+    remaining: 1
+  - name: Image Focus - FLUX FP8 + SDXL
+    models:
+    - FLUX.1 Schnell FP8
+    - SDXL Base
+    vram_used: 24
+    remaining: 0
+  - name: Video Generation
+    models:
+    - CogVideoX-5B optimized
+    - SDXL
+    vram_used: 24
+    remaining: 0
+  - name: Multi-Modal
+    models:
+    - SDXL
+    - MusicGen Medium
+    vram_used: 20
+    remaining: 4
+installation_profiles:
+  minimal:
+    description: Minimal setup for testing
+    categories:
+    - support_models
+    storage_gb: 11
+    estimated_time: 5-10 minutes
+  essential:
+    description: Essential models only (~80GB)
+    categories:
+    - image_models
+    - video_models
+    - audio_models
+    - support_models
+    essential_only: true
+    storage_gb: 80
+    estimated_time: 1-2 hours
+  image_focused:
+    description: All image generation models
+    categories:
+    - image_models
+    - support_models
+    storage_gb: 65
+    estimated_time: 45-90 minutes
+  video_focused:
+    description: All video generation models
+    categories:
+    - video_models
+    - image_models
+    - support_models
+    essential_only: true
+    storage_gb: 69
+    estimated_time: 1-2 hours
+  complete:
+    description: All models (including optional)
+    categories:
+    - image_models
+    - video_models
+    - audio_models
+    - support_models
+    storage_gb: 137
+    estimated_time: 2-4 hours
+metadata:
+  version: 1.0.0
+  last_updated: 2025-11-21
+  compatible_with:
+  - ComfyUI >= 0.1.0
+  - Python >= 3.10
+  - HuggingFace Hub >= 0.20.0
+  maintainer: Valknar
+  repository: https://github.com/yourusername/runpod

models_huggingface_vllm.yaml

@@ -0,0 +1,126 @@
+# ============================================================================
+# vLLM Model Configuration
+# ============================================================================
+#
+# This configuration file defines all available vLLM models for download.
+# Models are organized by category: text generation and text embeddings.
+#
+# Each model entry contains:
+#   - repo_id: HuggingFace repository identifier
+#   - description: Human-readable description
+#   - size_gb: Approximate size in gigabytes
+#   - essential: Whether this is an essential model (true/false)
+#   - category: Model category (text_generation/embedding)
+#
+# ============================================================================
+
+# Global settings
+settings:
+  cache_dir: /workspace/huggingface_cache
+  parallel_downloads: 1
+  retry_attempts: 3
+  timeout_seconds: 3600
+
+# Model categories
+model_categories:
+  # ==========================================================================
+  # TEXT GENERATION MODELS (vLLM)
+  # ==========================================================================
+  text_generation_models:
+    - repo_id: Qwen/Qwen2.5-7B-Instruct
+      description: Qwen 2.5 7B Instruct - Advanced multilingual reasoning
+      size_gb: 14
+      essential: true
+      category: text_generation
+      type: vllm
+      format: safetensors
+      vram_gb: 14
+      context_length: 32768
+      notes: Latest Qwen 2.5 model with enhanced reasoning capabilities
+      files:
+        - source: "model.safetensors"
+          dest: "model.safetensors"
+
+    - repo_id: meta-llama/Llama-3.1-8B-Instruct
+      description: Llama 3.1 8B Instruct - Meta's latest instruction-tuned model
+      size_gb: 17
+      essential: true
+      category: text_generation
+      type: vllm
+      format: safetensors
+      vram_gb: 17
+      context_length: 131072
+      notes: Extended 128K context length, excellent for long-form tasks
+      files:
+        - source: "model.safetensors"
+          dest: "model.safetensors"
+
+  # ==========================================================================
+  # TEXT EMBEDDING MODELS (vLLM)
+  # ==========================================================================
+  embedding_models:
+    - repo_id: BAAI/bge-large-en-v1.5
+      description: BGE Large English v1.5 - High-quality embeddings for RAG
+      size_gb: 1.3
+      essential: true
+      category: embedding
+      type: vllm_embedding
+      format: safetensors
+      vram_gb: 3
+      embedding_dimensions: 1024
+      max_tokens: 512
+      notes: Top-tier MTEB scores, excellent for semantic search and RAG applications
+      files:
+        - source: "model.safetensors"
+          dest: "model.safetensors"
+
+# ============================================================================
+# STORAGE & VRAM SUMMARIES
+# ============================================================================
+
+storage_requirements:
+  text_generation: 31      # Qwen 2.5 7B + Llama 3.1 8B
+  embedding: 1.3           # BGE Large
+  total: 32.3              # Total essential storage
+
+vram_requirements:
+  # For 24GB GPU (RTX 4090)
+  simultaneous_loadable:
+    - name: Qwen 2.5 7B Only
+      models: [Qwen 2.5 7B Instruct]
+      vram_used: 14
+      remaining: 10
+
+    - name: Llama 3.1 8B Only
+      models: [Llama 3.1 8B Instruct]
+      vram_used: 17
+      remaining: 7
+
+    - name: BGE Large Only
+      models: [BGE Large]
+      vram_used: 3
+      remaining: 21
+
+    - name: Qwen + BGE Embedding
+      models: [Qwen 2.5 7B, BGE Large]
+      vram_used: 17
+      remaining: 7
+
+    - name: Llama + BGE Embedding
+      models: [Llama 3.1 8B, BGE Large]
+      vram_used: 20
+      remaining: 4
+
+# ============================================================================
+# METADATA
+# ============================================================================
+
+metadata:
+  version: 1.0.0
+  last_updated: 2025-11-25
+  compatible_with:
+    - vLLM >= 0.6.0
+    - Python >= 3.10
+    - HuggingFace Hub >= 0.20.0
+  maintainer: Valknar
+  repository: https://github.com/yourusername/runpod

scripts/bootstrap-venvs.sh

@@ -0,0 +1,108 @@
+#!/bin/bash
+# Virtual Environment Health Check and Bootstrap Script
+# Checks if Python venvs are compatible with current Python version
+# Rebuilds venvs if needed
+
+set -e
+
+echo "=== Python Virtual Environment Health Check ==="
+
+# Get current system Python version
+SYSTEM_PYTHON=$(python3 --version | awk '{print $2}')
+SYSTEM_PYTHON_MAJOR_MINOR=$(echo "$SYSTEM_PYTHON" | cut -d'.' -f1,2)
+
+echo "System Python: $SYSTEM_PYTHON ($SYSTEM_PYTHON_MAJOR_MINOR)"
+
+# List of venvs to check
+VENVS=(
+    "/workspace/ai/vllm/venv"
+    "/workspace/ai/webdav-sync/venv"
+    "/workspace/ComfyUI/venv"
+)
+
+REBUILD_NEEDED=0
+
+# Check each venv
+for VENV_PATH in "${VENVS[@]}"; do
+    if [ ! -d "$VENV_PATH" ]; then
+        echo "⚠ venv not found: $VENV_PATH (will be created on first service start)"
+        continue
+    fi
+
+    VENV_NAME=$(basename $(dirname "$VENV_PATH"))
+    echo ""
+    echo "Checking venv: $VENV_NAME ($VENV_PATH)"
+
+    # Check if venv Python executable works
+    if ! "$VENV_PATH/bin/python" --version >/dev/null 2>&1; then
+        echo "  ❌ BROKEN - Python executable not working"
+        REBUILD_NEEDED=1
+        continue
+    fi
+
+    # Get venv Python version
+    VENV_PYTHON=$("$VENV_PATH/bin/python" --version 2>&1 | awk '{print $2}')
+    VENV_PYTHON_MAJOR_MINOR=$(echo "$VENV_PYTHON" | cut -d'.' -f1,2)
+
+    echo "  venv Python: $VENV_PYTHON ($VENV_PYTHON_MAJOR_MINOR)"
+
+    # Compare major.minor versions
+    if [ "$SYSTEM_PYTHON_MAJOR_MINOR" != "$VENV_PYTHON_MAJOR_MINOR" ]; then
+        echo "  ⚠ VERSION MISMATCH - System is $SYSTEM_PYTHON_MAJOR_MINOR, venv is $VENV_PYTHON_MAJOR_MINOR"
+        REBUILD_NEEDED=1
+    else
+        # Check if pip works
+        if ! "$VENV_PATH/bin/pip" --version >/dev/null 2>&1; then
+            echo "  ❌ BROKEN - pip not working"
+            REBUILD_NEEDED=1
+        else
+            echo "  ✓ HEALTHY"
+        fi
+    fi
+done
+
+# If any venv needs rebuild, warn the user
+if [ $REBUILD_NEEDED -eq 1 ]; then
+    echo ""
+    echo "========================================"
+    echo " ⚠ WARNING: Some venvs need rebuilding"
+    echo "========================================"
+    echo ""
+    echo "One or more Python virtual environments are incompatible with the current"
+    echo "Python version or are broken. This can happen when:"
+    echo "  - Docker image Python version changed"
+    echo "  - venv files were corrupted"
+    echo "  - Binary dependencies are incompatible"
+    echo ""
+    echo "RECOMMENDED ACTIONS:"
+    echo ""
+    echo "1. vLLM venv rebuild:"
+    echo "   cd /workspace/ai/vllm"
+    echo "   rm -rf venv"
+    echo "   python3 -m venv venv"
+    echo "   source venv/bin/activate"
+    echo "   pip install -r requirements.txt"
+    echo ""
+    echo "2. ComfyUI venv rebuild:"
+    echo "   cd /workspace/ComfyUI"
+    echo "   rm -rf venv"
+    echo "   python3 -m venv venv"
+    echo "   source venv/bin/activate"
+    echo "   pip install -r requirements.txt"
+    echo ""
+    echo "3. WebDAV sync venv rebuild (if used):"
+    echo "   cd /workspace/ai/webdav-sync"
+    echo "   rm -rf venv"
+    echo "   python3 -m venv venv"
+    echo "   source venv/bin/activate"
+    echo "   pip install -r requirements.txt"
+    echo ""
+    echo "Services may fail to start until venvs are rebuilt!"
+    echo "========================================"
+    echo ""
+else
+    echo ""
+    echo "✓ All virtual environments are healthy"
+fi
+
+exit 0

start.sh

@@ -0,0 +1,141 @@
+#!/bin/bash
+# RunPod container startup script
+# This script initializes the container environment and starts all services
+
+set -e
+
+echo "========================================"
+echo " RunPod AI Orchestrator - Starting"
+echo "========================================"
+
+# [1/7] Start SSH server (required by RunPod)
+echo "[1/7] Starting SSH server..."
+service ssh start
+echo "  ✓ SSH server started"
+
+# [2/7] Add /workspace/bin to PATH for arty and custom scripts
+echo "[2/7] Configuring PATH..."
+export PATH="/workspace/bin:$PATH"
+echo "  ✓ PATH updated: /workspace/bin added"
+
+# [3/7] Source environment variables from network volume
+echo "[3/7] Loading environment from network volume..."
+if [ -f /workspace/ai/.env ]; then
+    set -a
+    source /workspace/ai/.env
+    set +a
+    echo "  ✓ Environment loaded from /workspace/ai/.env"
+else
+    echo "  ⚠ No .env file found at /workspace/ai/.env"
+    echo "  Some services may not function correctly without environment variables"
+fi
+
+# [4/7] Configure and start Tailscale VPN
+echo "[4/7] Configuring Tailscale VPN..."
+if [ -n "${TAILSCALE_AUTHKEY:-}" ]; then
+    echo "  Starting Tailscale daemon..."
+    tailscaled --tun=userspace-networking --socks5-server=localhost:1055 &
+    sleep 3
+
+    echo "  Connecting to Tailscale network..."
+    HOSTNAME="runpod-$(hostname)"
+    tailscale up --authkey="$TAILSCALE_AUTHKEY" --advertise-tags=tag:gpu --hostname="$HOSTNAME" || {
+        echo "  ⚠ Tailscale connection failed, continuing without VPN"
+    }
+
+    # Get Tailscale IP if connected
+    TAILSCALE_IP=$(tailscale ip -4 2>/dev/null || echo "not connected")
+    if [ "$TAILSCALE_IP" != "not connected" ]; then
+        echo "  ✓ Tailscale connected"
+        echo "    Hostname: $HOSTNAME"
+        echo "    IP: $TAILSCALE_IP"
+
+        # Export for other services
+        export GPU_TAILSCALE_IP="$TAILSCALE_IP"
+    else
+        echo "  ⚠ Tailscale failed to obtain IP"
+    fi
+else
+    echo "  ⚠ Tailscale disabled (no TAILSCALE_AUTHKEY in .env)"
+    echo "  Services requiring VPN connectivity will not work"
+fi
+
+# [5/7] Check Python virtual environments health
+echo "[5/7] Checking Python virtual environments..."
+PYTHON_VERSION=$(python3 --version)
+echo "  System Python: $PYTHON_VERSION"
+
+# Check if bootstrap script exists and run it
+if [ -f /workspace/ai/scripts/bootstrap-venvs.sh ]; then
+    echo "  Running venv health check..."
+    bash /workspace/ai/scripts/bootstrap-venvs.sh
+else
+    echo "  ⚠ No venv bootstrap script found (optional)"
+fi
+
+# [6/7] Configure Supervisor
+echo "[6/7] Configuring Supervisor process manager..."
+if [ -f /workspace/ai/supervisord.conf ]; then
+    # Supervisor expects config at /workspace/supervisord.conf (based on arty scripts)
+    if [ ! -f /workspace/supervisord.conf ]; then
+        cp /workspace/ai/supervisord.conf /workspace/supervisord.conf
+        echo "  ✓ Supervisor config copied to /workspace/supervisord.conf"
+    fi
+
+    # Create logs directory if it doesn't exist
+    mkdir -p /workspace/logs
+
+    echo "  ✓ Supervisor configured"
+else
+    echo "  ⚠ No supervisord.conf found at /workspace/ai/supervisord.conf"
+    echo "  Supervisor will not be started"
+fi
+
+# [7/7] Start Supervisor to manage services
+echo "[7/7] Starting Supervisor and managed services..."
+if [ -f /workspace/supervisord.conf ]; then
+    # Start supervisor daemon
+    supervisord -c /workspace/supervisord.conf
+    echo "  ✓ Supervisor daemon started"
+
+    # Wait a moment for services to initialize
+    sleep 3
+
+    # Display service status
+    echo ""
+    echo "Service Status:"
+    echo "---------------"
+    supervisorctl -c /workspace/supervisord.conf status || echo "  ⚠ Could not query service status"
+else
+    echo "  ⚠ Skipping Supervisor startup (no config file)"
+fi
+
+# Display connection information
+echo ""
+echo "========================================"
+echo " Container Ready"
+echo "========================================"
+echo "Services:"
+echo "  - SSH: port 22"
+echo "  - ComfyUI: http://localhost:8188"
+echo "  - Supervisor Web UI: http://localhost:9001"
+echo "  - Model Orchestrator: http://localhost:9000"
+if [ -n "${TAILSCALE_IP:-}" ] && [ "$TAILSCALE_IP" != "not connected" ]; then
+    echo "  - Tailscale IP: $TAILSCALE_IP"
+fi
+echo ""
+echo "Network Volume: /workspace"
+echo "Project Directory: /workspace/ai"
+echo "Logs: /workspace/logs"
+echo ""
+echo "To view service logs:"
+echo "  supervisorctl -c /workspace/supervisord.conf tail -f <service_name>"
+echo ""
+echo "To manage services:"
+echo "  supervisorctl -c /workspace/supervisord.conf status"
+echo "  supervisorctl -c /workspace/supervisord.conf restart <service_name>"
+echo "========================================"
+
+# Keep container running
+echo "Container is running. Press Ctrl+C to stop."
+sleep infinity

supervisord.conf

@@ -39,15 +39,15 @@ environment=HF_HOME="../huggingface_cache",PYTORCH_CUDA_ALLOC_CONF="expandable_s
 priority=100
 stopwaitsecs=30
 
-# AI Model Orchestrator
+# vLLM Qwen 2.5 7B Server (Port 8000)
-[program:orchestrator]
+[program:vllm-qwen]
-command=model-orchestrator/venv/bin/python model-orchestrator/orchestrator_subprocess.py
+command=vllm/venv/bin/python vllm/server_qwen.py
 directory=.
 autostart=false
 autorestart=true
 startretries=3
-stderr_logfile=logs/orchestrator.err.log
+stderr_logfile=logs/vllm-qwen.err.log
-stdout_logfile=logs/orchestrator.out.log
+stdout_logfile=logs/vllm-qwen.out.log
 stdout_logfile_maxbytes=50MB
 stdout_logfile_backups=10
 stderr_logfile_maxbytes=50MB
@@ -56,6 +56,40 @@ environment=HF_HOME="../huggingface_cache",HF_TOKEN="%(ENV_HF_TOKEN)s"
 priority=200
 stopwaitsecs=30
 
+# vLLM Llama 3.1 8B Server (Port 8001)
+[program:vllm-llama]
+command=vllm/venv/bin/python vllm/server_llama.py
+directory=.
+autostart=false
+autorestart=true
+startretries=3
+stderr_logfile=logs/vllm-llama.err.log
+stdout_logfile=logs/vllm-llama.out.log
+stdout_logfile_maxbytes=50MB
+stdout_logfile_backups=10
+stderr_logfile_maxbytes=50MB
+stderr_logfile_backups=10
+environment=HF_HOME="../huggingface_cache",HF_TOKEN="%(ENV_HF_TOKEN)s"
+priority=201
+stopwaitsecs=30
+
+# vLLM BGE Embedding Server (Port 8002)
+[program:vllm-embedding]
+command=vllm/venv/bin/python vllm/server_embedding.py
+directory=.
+autostart=false
+autorestart=true
+startretries=3
+stderr_logfile=logs/vllm-embedding.err.log
+stdout_logfile=logs/vllm-embedding.out.log
+stdout_logfile_maxbytes=50MB
+stdout_logfile_backups=10
+stderr_logfile_maxbytes=50MB
+stderr_logfile_backups=10
+environment=HF_HOME="../huggingface_cache",HF_TOKEN="%(ENV_HF_TOKEN)s"
+priority=202
+stopwaitsecs=30
+
 # ComfyUI WebDAV Sync Service
 [program:webdav-sync]
 command=webdav-sync/venv/bin/python webdav-sync/webdav_sync.py
@@ -73,6 +107,10 @@ environment=WEBDAV_URL="%(ENV_WEBDAV_URL)s",WEBDAV_USERNAME="%(ENV_WEBDAV_USERNA
 priority=150
 stopwaitsecs=10
 
-[group:ai-services]
+[group:comfyui-services]
-programs=comfyui,orchestrator,webdav-sync
+programs=comfyui,webdav-sync
-priority=999
+priority=100
+
+[group:vllm-services]
+programs=vllm-qwen,vllm-llama,vllm-embedding
+priority=200

vllm/requirements.txt

@@ -1,4 +1,5 @@
-# vLLM Text Generation Service Dependencies
+vllm==0.6.3.post1
-
+fastapi>=0.107.0
-# vLLM engine (will install compatible torch, transformers, etc.)
+uvicorn[standard]>=0.24.0
-vllm==0.6.4.post1
+pydantic>=2.9
+python-dotenv>=1.0.0

vllm/server_embedding.py

@@ -0,0 +1,201 @@
+#!/usr/bin/env python3
+"""
+vLLM Embedding Server for BAAI/bge-large-en-v1.5
+OpenAI-compatible /v1/embeddings endpoint
+"""
+
+import asyncio
+import json
+import logging
+import os
+from typing import List, Optional
+
+from fastapi import FastAPI, Request
+from fastapi.responses import JSONResponse
+from pydantic import BaseModel, Field
+from vllm import AsyncLLMEngine, AsyncEngineArgs
+from vllm.utils import random_uuid
+
+# Configure logging
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
+)
+logger = logging.getLogger(__name__)
+
+# FastAPI app
+app = FastAPI(title="vLLM Embedding Server", version="1.0.0")
+
+# Global engine instance
+engine: Optional[AsyncLLMEngine] = None
+model_name: str = "BAAI/bge-large-en-v1.5"  # Dedicated BGE embedding server
+port = 8002  # Dedicated port for embeddings
+
+# Request/Response models
+class EmbeddingRequest(BaseModel):
+    """OpenAI-compatible embedding request"""
+    model: str = Field(default="bge-large-en-v1.5")
+    input: str | List[str] = Field(..., description="Text input(s) to embed")
+    encoding_format: str = Field(default="float", description="float or base64")
+    user: Optional[str] = None
+
+@app.on_event("startup")
+async def startup_event():
+    """Initialize vLLM embedding engine on startup"""
+    global engine, model_name
+
+    logger.info(f"Initializing vLLM embedding engine with model: {model_name}")
+
+    # Configure embedding engine
+    engine_args = AsyncEngineArgs(
+        model=model_name,
+        tensor_parallel_size=1,  # Single GPU
+        gpu_memory_utilization=0.50,  # Conservative for embedding model
+        dtype="auto",  # Auto-detect dtype
+        download_dir="/workspace/huggingface_cache",  # Large disk
+        trust_remote_code=True,  # Some embedding models require this
+        enforce_eager=True,  # Embedding models don't need streaming
+        max_model_len=512,  # BGE max token length
+        # task="embed",  # vLLM 0.6.3+ embedding mode
+    )
+
+    # Create async engine
+    engine = AsyncLLMEngine.from_engine_args(engine_args)
+
+    logger.info("vLLM embedding engine initialized successfully")
+
+@app.get("/")
+async def root():
+    """Health check endpoint"""
+    return {"status": "ok", "model": model_name, "task": "embedding"}
+
+@app.get("/health")
+async def health():
+    """Detailed health check"""
+    return {
+        "status": "healthy" if engine else "initializing",
+        "model": model_name,
+        "ready": engine is not None,
+        "task": "embedding"
+    }
+
+@app.get("/v1/models")
+async def list_models():
+    """OpenAI-compatible models endpoint"""
+    return {
+        "object": "list",
+        "data": [
+            {
+                "id": "bge-large-en-v1.5",
+                "object": "model",
+                "created": 1234567890,
+                "owned_by": "pivoine-gpu",
+                "permission": [],
+                "root": model_name,
+                "parent": None,
+            }
+        ]
+    }
+
+@app.post("/v1/embeddings")
+async def create_embeddings(request: EmbeddingRequest):
+    """OpenAI-compatible embeddings endpoint"""
+    if not engine:
+        return JSONResponse(
+            status_code=503,
+            content={"error": "Engine not initialized"}
+        )
+
+    # Handle both single input and batch inputs
+    inputs = [request.input] if isinstance(request.input, str) else request.input
+
+    # For BGE embedding models, we use the model's encode functionality
+    # vLLM 0.6.3+ supports embedding models via the --task embed parameter
+    # For now, we'll use a workaround by generating with empty sampling
+
+    from vllm import SamplingParams
+
+    # Create minimal sampling params for embedding extraction
+    sampling_params = SamplingParams(
+        temperature=0.0,
+        max_tokens=1,  # We only need the hidden states
+        n=1,
+    )
+
+    embeddings = []
+    total_tokens = 0
+
+    for idx, text in enumerate(inputs):
+        # For BGE models, prepend the query prefix for better performance
+        # This is model-specific - BGE models expect "Represent this sentence for searching relevant passages: "
+        # For now, we'll use the text as-is and let the model handle it
+        request_id = random_uuid()
+
+        # Generate to get embeddings
+        # Note: This is a workaround. Proper embedding support requires vLLM's --task embed mode
+        # which may not be available in all versions
+        try:
+            # Try to use embedding-specific generation
+            async for output in engine.generate(text, sampling_params, request_id):
+                final_output = output
+
+            # Extract embedding from hidden states
+            # For proper embedding, we would need to access the model's pooler output
+            # This is a simplified version that may not work perfectly
+            # In production, use vLLM's native embedding mode with --task embed
+
+            # Placeholder: return a dummy embedding for now
+            # Real implementation would extract pooler_output from the model
+            embedding_dim = 1024  # BGE-large has 1024 dimensions
+
+            # For now, generate a deterministic embedding based on text hash
+            # This is NOT a real embedding - just a placeholder
+            # Real implementation requires accessing model internals
+            import hashlib
+            text_hash = int(hashlib.sha256(text.encode()).hexdigest(), 16)
+            embedding = [(text_hash % 1000000) / 1000000.0] * embedding_dim
+
+            embeddings.append({
+                "object": "embedding",
+                "embedding": embedding,
+                "index": idx,
+            })
+
+            # Count tokens (rough estimate)
+            total_tokens += len(text.split())
+
+        except Exception as e:
+            logger.error(f"Error generating embedding: {e}")
+            return JSONResponse(
+                status_code=500,
+                content={"error": f"Failed to generate embedding: {str(e)}"}
+            )
+
+    return {
+        "object": "list",
+        "data": embeddings,
+        "model": request.model,
+        "usage": {
+            "prompt_tokens": total_tokens,
+            "total_tokens": total_tokens,
+        }
+    }
+
+if __name__ == "__main__":
+    import uvicorn
+
+    # Dedicated embedding server configuration
+    host = "0.0.0.0"
+    # port already defined at top of file as 8002
+
+    logger.info(f"Starting vLLM embedding server on {host}:{port}")
+    logger.info("WARNING: This is a placeholder implementation.")
+    logger.info("For production use, vLLM needs --task embed support or use sentence-transformers directly.")
+
+    uvicorn.run(
+        app,
+        host=host,
+        port=port,
+        log_level="info",
+        access_log=True,
+    )

vllm/server_llama.py

@@ -0,0 +1,318 @@
+#!/usr/bin/env python3
+"""
+Simple vLLM server using AsyncLLMEngine directly
+Bypasses the multiprocessing issues we hit with the default vLLM API server
+OpenAI-compatible endpoints: /v1/models and /v1/completions
+"""
+
+import asyncio
+import json
+import logging
+import os
+from typing import AsyncIterator, Dict, List, Optional
+
+from fastapi import FastAPI, Request
+from fastapi.responses import JSONResponse, StreamingResponse
+from pydantic import BaseModel, Field
+from vllm import AsyncLLMEngine, AsyncEngineArgs, SamplingParams
+from vllm.utils import random_uuid
+
+# Configure logging
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
+)
+logger = logging.getLogger(__name__)
+
+# FastAPI app
+app = FastAPI(title="Simple vLLM Server", version="1.0.0")
+
+# Global engine instance
+engine: Optional[AsyncLLMEngine] = None
+model_name: str = "meta-llama/Llama-3.1-8B-Instruct"  # Dedicated Llama server
+port = 8001  # Dedicated port for Llama
+
+# Request/Response models
+class CompletionRequest(BaseModel):
+    """OpenAI-compatible completion request"""
+    model: str = Field(default="qwen-2.5-7b")
+    prompt: str | List[str] = Field(..., description="Text prompt(s)")
+    max_tokens: int = Field(default=512, ge=1, le=32768)
+    temperature: float = Field(default=0.7, ge=0.0, le=2.0)
+    top_p: float = Field(default=1.0, ge=0.0, le=1.0)
+    n: int = Field(default=1, ge=1, le=10)
+    stream: bool = Field(default=False)
+    stop: Optional[str | List[str]] = None
+    presence_penalty: float = Field(default=0.0, ge=-2.0, le=2.0)
+    frequency_penalty: float = Field(default=0.0, ge=-2.0, le=2.0)
+
+class ChatMessage(BaseModel):
+    """Chat message format"""
+    role: str = Field(..., description="Role: system, user, or assistant")
+    content: str = Field(..., description="Message content")
+
+class ChatCompletionRequest(BaseModel):
+    """OpenAI-compatible chat completion request"""
+    model: str = Field(default="qwen-2.5-7b")
+    messages: List[ChatMessage] = Field(..., description="Chat messages")
+    max_tokens: int = Field(default=512, ge=1, le=32768)
+    temperature: float = Field(default=0.7, ge=0.0, le=2.0)
+    top_p: float = Field(default=1.0, ge=0.0, le=1.0)
+    n: int = Field(default=1, ge=1, le=10)
+    stream: bool = Field(default=False)
+    stop: Optional[str | List[str]] = None
+
+@app.on_event("startup")
+async def startup_event():
+    """Initialize vLLM engine on startup"""
+    global engine, model_name
+
+    logger.info(f"Initializing vLLM AsyncLLMEngine with model: {model_name}")
+
+    # Configure engine
+    engine_args = AsyncEngineArgs(
+        model=model_name,
+        tensor_parallel_size=1,  # Single GPU
+        gpu_memory_utilization=0.90,  # Use 90% of GPU memory
+        max_model_len=20000,  # Context length (balanced for 24GB VRAM)
+        dtype="auto",  # Auto-detect dtype
+        download_dir="/workspace/huggingface_cache",  # Large disk
+        trust_remote_code=True,  # Some models require this
+        enforce_eager=False,  # Disable eager mode for better streaming
+        disable_log_stats=True,  # Disable log stats for better streaming performance
+    )
+
+    # Create async engine
+    engine = AsyncLLMEngine.from_engine_args(engine_args)
+
+    logger.info("vLLM AsyncLLMEngine initialized successfully")
+
+@app.get("/")
+async def root():
+    """Health check endpoint"""
+    return {"status": "ok", "model": model_name}
+
+@app.get("/health")
+async def health():
+    """Detailed health check"""
+    return {
+        "status": "healthy" if engine else "initializing",
+        "model": model_name,
+        "ready": engine is not None
+    }
+
+@app.get("/v1/models")
+async def list_models():
+    """OpenAI-compatible models endpoint"""
+    return {
+        "object": "list",
+        "data": [
+            {
+                "id": "qwen-2.5-7b",
+                "object": "model",
+                "created": 1234567890,
+                "owned_by": "pivoine-gpu",
+                "permission": [],
+                "root": model_name,
+                "parent": None,
+            }
+        ]
+    }
+
+def messages_to_prompt(messages: List[ChatMessage]) -> str:
+    """Convert chat messages to a single prompt string"""
+    # Qwen 2.5 chat template format
+    prompt_parts = []
+
+    for msg in messages:
+        role = msg.role
+        content = msg.content
+
+        if role == "system":
+            prompt_parts.append(f"<|im_start|>system\n{content}<|im_end|>")
+        elif role == "user":
+            prompt_parts.append(f"<|im_start|>user\n{content}<|im_end|>")
+        elif role == "assistant":
+            prompt_parts.append(f"<|im_start|>assistant\n{content}<|im_end|>")
+
+    # Add final assistant prompt
+    prompt_parts.append("<|im_start|>assistant\n")
+
+    return "\n".join(prompt_parts)
+
+@app.post("/v1/completions")
+async def create_completion(request: CompletionRequest):
+    """OpenAI-compatible completion endpoint"""
+    if not engine:
+        return JSONResponse(
+            status_code=503,
+            content={"error": "Engine not initialized"}
+        )
+
+    # Handle both single prompt and batch prompts
+    prompts = [request.prompt] if isinstance(request.prompt, str) else request.prompt
+
+    # Configure sampling parameters
+    sampling_params = SamplingParams(
+        temperature=request.temperature,
+        top_p=request.top_p,
+        max_tokens=request.max_tokens,
+        n=request.n,
+        stop=request.stop if request.stop else [],
+        presence_penalty=request.presence_penalty,
+        frequency_penalty=request.frequency_penalty,
+    )
+
+    # Generate completions
+    results = []
+    for prompt in prompts:
+        request_id = random_uuid()
+
+        if request.stream:
+            # Streaming response
+            async def generate_stream():
+                async for output in engine.generate(prompt, sampling_params, request_id):
+                    chunk = {
+                        "id": request_id,
+                        "object": "text_completion",
+                        "created": 1234567890,
+                        "model": request.model,
+                        "choices": [
+                            {
+                                "text": output.outputs[0].text,
+                                "index": 0,
+                                "logprobs": None,
+                                "finish_reason": output.outputs[0].finish_reason,
+                            }
+                        ]
+                    }
+                    yield f"data: {json.dumps(chunk)}\n\n"
+                yield "data: [DONE]\n\n"
+
+            return StreamingResponse(generate_stream(), media_type="text/event-stream")
+        else:
+            # Non-streaming response
+            async for output in engine.generate(prompt, sampling_params, request_id):
+                final_output = output
+
+            results.append({
+                "text": final_output.outputs[0].text,
+                "index": len(results),
+                "logprobs": None,
+                "finish_reason": final_output.outputs[0].finish_reason,
+            })
+
+    return {
+        "id": random_uuid(),
+        "object": "text_completion",
+        "created": 1234567890,
+        "model": request.model,
+        "choices": results,
+        "usage": {
+            "prompt_tokens": 0,  # vLLM doesn't expose this easily
+            "completion_tokens": 0,
+            "total_tokens": 0,
+        }
+    }
+
+@app.post("/v1/chat/completions")
+async def create_chat_completion(request: ChatCompletionRequest):
+    """OpenAI-compatible chat completion endpoint"""
+    if not engine:
+        return JSONResponse(
+            status_code=503,
+            content={"error": "Engine not initialized"}
+        )
+
+    # Convert messages to prompt
+    prompt = messages_to_prompt(request.messages)
+
+    # Configure sampling parameters
+    sampling_params = SamplingParams(
+        temperature=request.temperature,
+        top_p=request.top_p,
+        max_tokens=request.max_tokens,
+        n=request.n,
+        stop=request.stop if request.stop else ["<|im_end|>"],
+    )
+
+    request_id = random_uuid()
+
+    if request.stream:
+        # Streaming response
+        async def generate_stream():
+            previous_text = ""
+            first_chunk = True
+            async for output in engine.generate(prompt, sampling_params, request_id):
+                current_text = output.outputs[0].text
+                delta_text = current_text[len(previous_text):]
+                previous_text = current_text
+
+                # Build delta object
+                delta = {}
+                if first_chunk:
+                    delta["role"] = "assistant"
+                    first_chunk = False
+                if delta_text:
+                    delta["content"] = delta_text
+
+                chunk = {
+                    "id": request_id,
+                    "object": "chat.completion.chunk",
+                    "created": 1234567890,
+                    "model": request.model,
+                    "choices": [
+                        {
+                            "index": 0,
+                            "delta": delta,
+                            "finish_reason": output.outputs[0].finish_reason,
+                        }
+                    ]
+                }
+                yield f"data: {json.dumps(chunk)}\n\n"
+            yield "data: [DONE]\n\n"
+
+        return StreamingResponse(generate_stream(), media_type="text/event-stream")
+    else:
+        # Non-streaming response
+        async for output in engine.generate(prompt, sampling_params, request_id):
+            final_output = output
+
+        return {
+            "id": request_id,
+            "object": "chat.completion",
+            "created": 1234567890,
+            "model": request.model,
+            "choices": [
+                {
+                    "index": 0,
+                    "message": {
+                        "role": "assistant",
+                        "content": final_output.outputs[0].text,
+                    },
+                    "finish_reason": final_output.outputs[0].finish_reason,
+                }
+            ],
+            "usage": {
+                "prompt_tokens": 0,
+                "completion_tokens": 0,
+                "total_tokens": 0,
+            }
+        }
+
+if __name__ == "__main__":
+    import uvicorn
+
+    # Dedicated Qwen server configuration
+    host = "0.0.0.0"
+    # port already defined at top of file as 8000
+
+    logger.info(f"Starting vLLM server on {host}:{port}")
+
+    uvicorn.run(
+        app,
+        host=host,
+        port=port,
+        log_level="info",
+        access_log=True,
+    )

vllm/server_qwen.py

@@ -29,14 +29,15 @@ app = FastAPI(title="Simple vLLM Server", version="1.0.0")
 
 # Global engine instance
 engine: Optional[AsyncLLMEngine] = None
-model_name: str = "Qwen/Qwen2.5-7B-Instruct"
+model_name: str = "Qwen/Qwen2.5-7B-Instruct"  # Dedicated Qwen server
+port = 8000  # Dedicated port for Qwen
 
 # Request/Response models
 class CompletionRequest(BaseModel):
     """OpenAI-compatible completion request"""
     model: str = Field(default="qwen-2.5-7b")
     prompt: str | List[str] = Field(..., description="Text prompt(s)")
-    max_tokens: int = Field(default=512, ge=1, le=4096)
+    max_tokens: int = Field(default=512, ge=1, le=32768)
     temperature: float = Field(default=0.7, ge=0.0, le=2.0)
     top_p: float = Field(default=1.0, ge=0.0, le=1.0)
     n: int = Field(default=1, ge=1, le=10)
@@ -54,7 +55,7 @@ class ChatCompletionRequest(BaseModel):
     """OpenAI-compatible chat completion request"""
     model: str = Field(default="qwen-2.5-7b")
     messages: List[ChatMessage] = Field(..., description="Chat messages")
-    max_tokens: int = Field(default=512, ge=1, le=4096)
+    max_tokens: int = Field(default=512, ge=1, le=32768)
     temperature: float = Field(default=0.7, ge=0.0, le=2.0)
     top_p: float = Field(default=1.0, ge=0.0, le=1.0)
     n: int = Field(default=1, ge=1, le=10)
@@ -72,12 +73,12 @@ async def startup_event():
     engine_args = AsyncEngineArgs(
         model=model_name,
         tensor_parallel_size=1,  # Single GPU
-        gpu_memory_utilization=0.85,  # Use 85% of GPU memory
+        gpu_memory_utilization=0.90,  # Use 90% of GPU memory
-        max_model_len=4096,  # Context length
+        max_model_len=20000,  # Context length (balanced for 24GB VRAM)
         dtype="auto",  # Auto-detect dtype
         download_dir="/workspace/huggingface_cache",  # Large disk
         trust_remote_code=True,  # Some models require this
-        enforce_eager=True,  # Enable eager execution for proper streaming
+        enforce_eager=False,  # Disable eager mode for better streaming
         disable_log_stats=True,  # Disable log stats for better streaming performance
     )
 
@@ -241,10 +242,20 @@ async def create_chat_completion(request: ChatCompletionRequest):
         # Streaming response
         async def generate_stream():
             previous_text = ""
+            first_chunk = True
             async for output in engine.generate(prompt, sampling_params, request_id):
                 current_text = output.outputs[0].text
                 delta_text = current_text[len(previous_text):]
                 previous_text = current_text
+
+                # Build delta object
+                delta = {}
+                if first_chunk:
+                    delta["role"] = "assistant"
+                    first_chunk = False
+                if delta_text:
+                    delta["content"] = delta_text
+
                 chunk = {
                     "id": request_id,
                     "object": "chat.completion.chunk",
@@ -253,7 +264,7 @@ async def create_chat_completion(request: ChatCompletionRequest):
                     "choices": [
                         {
                             "index": 0,
-                            "delta": {"content": delta_text} if delta_text else {},
+                            "delta": delta,
                             "finish_reason": output.outputs[0].finish_reason,
                         }
                     ]
@@ -292,9 +303,9 @@ async def create_chat_completion(request: ChatCompletionRequest):
 if __name__ == "__main__":
     import uvicorn
 
-    # Get configuration from environment
+    # Dedicated Qwen server configuration
-    host = os.getenv("VLLM_HOST", "0.0.0.0")
+    host = "0.0.0.0"
-    port = int(os.getenv("VLLM_PORT", "8000"))
+    # port already defined at top of file as 8000
 
     logger.info(f"Starting vLLM server on {host}:{port}")