What is a LoRA and how do I use one?

Question

Accepted Answer

LoRA (Low-Rank Adaptation) is a technique for adding new styles, characters, or concepts to AI models without retraining the entire model. LoRAs are small files (typically 10-200MB) that modify how a model generates images.

To use a LoRA in Stable Diffusion:
1. Download a LoRA file (.safetensors) from sites like Civitai
2. Place it in your LoRAs folder
3. Reference it in your prompt: `<lora:filename:0.7>`

The number (0.7) controls strength. Start around 0.5-0.7 and adjust based on results.