Image-to-LoRA V2 generates style LoRAs without training

View on Reddit

i2L V2 turns one or more reference images into a style LoRA in a single forward pass, eliminating the need for explicit training. The updated version is available as a model collection on GitHub.