What is a LoRA?
LoRA (Low-Rank Adaptation) is a technique for fine-tuning Stable Diffusion models on specific styles, characters, or concepts without training a full model. Think of it as a small add-on that teaches SD new visual concepts.
What Can LoRAs Do?
- Replicate specific art styles consistently
- Generate specific characters or people
- Add specific clothing, objects, or concepts
- Apply consistent aesthetic filters
Where to Find LoRAs
- Civitai: The largest LoRA repository with previews and ratings
- Hugging Face: Community models and research LoRAs
- GitHub: Open-source LoRA training scripts and models
How to Use LoRAs
In your prompt, add the LoRA trigger with a weight:
<lora:style_name:0.8> your prompt here
Tips
- Start with weight 0.7-0.8 and adjust
- Don't stack too many LoRAs — they can conflict
- Use PromptFrom to generate base prompts, then add your LoRA triggers