feat: add comprehensive negative embeddings support (SD 1.5, SDXL, Pony)

- Add 3 new embedding categories to models_civitai.yaml:
  - embeddings_sd15: 6 embeddings (BadDream, UnrealisticDream, badhandv4, EasyNegative, FastNegativeV2, BadNegAnatomyV1-neg)
  - embeddings_sdxl: 1 embedding (BadX v1.1)
  - embeddings_pony: 2 embeddings (zPDXL3, zPDXLxxx)
- Total storage: ~1.1 MB (9 embeddings)
- Add comprehensive embeddings documentation to NSFW README
- Include usage examples, compatibility notes, and syntax guide
- Document embedding weights and recommended combinations
This commit is contained in:
2025-11-23 19:39:18 +01:00
parent 68d3606cab
commit 5770563d9a
2 changed files with 236 additions and 0 deletions

View File

@@ -260,6 +260,98 @@ Use `nsfw-ultimate-upscale-production-v1.json` for professional 2x upscaling wit
### 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

View File

@@ -120,6 +120,150 @@ model_categories:
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"
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"
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"
recommended_settings:
strength: "1.0-2.0"
notes: "ONLY for Pony Diffusion models. Enables explicit NSFW content generation."
# ==========================================================================
# LORA MODELS (Style Enhancement & Detail)
# ==========================================================================