LoRA Models
Style and effect modifiers.
What is LoRA
LoRA (Low-Rank Adaptation) modifies base models without changing the entire model.
Popular LoRA Types
Style LoRA
| LoRA | Style |
|---|---|
| anime-style | Anime aesthetic |
| pixel-art | Pixel art |
| watercolor | Watercolor painting |
| oil-painting | Oil painting |
Character LoRA
- Specific character appearance
- Consistent character across images
- Use with trigger words
Concept LoRA
- Custom concepts
- Special effects
- Unique styles
Download Sources
Civitai
Primary source for LoRA models:
- LoRA category
- Search by style, character, concept
Hugging Face
- Some official LoRAs available
Installation
ComfyUI/models/loras/
anime-style-v1.safetensors
character-name-v1.safetensorsUsage
Basic
Node: LoraLoader
Select: Your LoRA
Strength: 0.5-1.0In Prompt
bash
# Some LoRAs need trigger words
trigger-word, another-triggerParameters
| Parameter | Value | Effect |
|---|---|---|
| Strength | 0.5-0.8 | Balanced |
| Strength | 0.8-1.0 | Strong effect |
| Strength | 0.3-0.5 | Subtle |
SDXL LoRA
SDXL requires SDXL-compatible LoRAs:
ComfyUI/models/loras/
sdxl-compatible-lora.safetensorsTips
- Check trigger words: Read model description
- Start low: 0.5-0.7 for new LoRA
- Combine: Use multiple LoRAs
Next Steps
- ControlNet Models - Structure control
- Embedding - Text embeddings