Regi Ellis
e915afb65f
- Add .github/workflows/publish_action.yml for Comfy registry - Update README: ComfyUI Manager install (recommended), Beta badge + warning - Minor docs tweaks
2.7 KiB
2.7 KiB
IllustriousMasterModel
Loads your AI art model. This is your starting point - pick your model and you're ready to create!
What does it do?
- Loads Illustrious XL and compatible checkpoint files
- Automatically configures optimal settings for the model
- Provides model, CLIP, and VAE outputs for your workflow
- Detects model version (EPS/VPred) for best quality
Inputs
- ckpt_name: Your model file (Illustrious XL, Pony, etc.)
- vae_name: VAE file (usually "Baked VAE" is best)
- clip_name: CLIP model (usually "ViT-L" or leave default)
- clip_strength: How strongly CLIP influences (1.0 = normal)
- optimization: Memory/speed settings (auto is usually best)
Outputs
- MODEL: The main AI model for sampling
- CLIP: Text understanding model for prompts
- VAE: Image encoder/decoder
- model_info: Details about loaded model
Artist Tips
- Illustrious XL v0.1 models: Great for anime/manga style
- NovelAI models: Excellent for detailed illustrations
- Pony models: Good for varied styles and characters
- Always use the VAE that came with your model for best colors
- Higher CLIP strength = more literal prompt following
Choosing Models
For Anime/Manga:
- Illustrious-XL-v0.1
- Any Illustrious-based merges
- NovelAI-based models
For Semi-Realistic:
- Illustrious merged with realistic
- Pony-based merges
For Artistic Styles:
- Models trained on specific artists
- Style-focused checkpoints
Memory Optimization
- Auto: Let ComfyUI decide (recommended)
- Low VRAM: For 6-8GB cards
- Normal: For 8-12GB cards
- High: For 12GB+ cards
- CPU: Emergency mode (very slow)
Common Issues & Fixes
"Out of Memory":
- Switch to "Low VRAM" mode
- Reduce batch size
- Use smaller resolution
"Wrong Colors":
- Check VAE selection
- Use "Baked VAE" if available
- Try different VAE file
"Model Not Found":
- Check models/checkpoints folder
- Refresh model list
- Ensure .safetensors extension
Example Workflows
Basic Setup:
- Select your Illustrious model
- Keep VAE on "Baked VAE"
- Connect to samplers and encoders
Model Comparison:
- Load multiple MasterModel nodes
- Use same prompt/seed
- Compare different models side-by-side
Memory-Conscious:
- Set optimization to "Low VRAM"
- Use smaller resolutions initially
- Upscale only final selections
Pro Tips
- Download models from Civitai or Hugging Face
- Keep models organized in subfolders
- Test new models with same seed for consistency
- Some models need specific negative prompts
- Read model cards for optimal settings
Next steps: IllustriousPrompt for text input For sampling: IllustriousKSamplerPro