Edit: For some reason this lora downloads as a .ico instead of .safetensors. I have no idea why. You'll have to manually rename it. Sorry about the inconvenience.Does what it says on the tin. Probably only works with "fp8 e4m3fn" Flux precision.<s>You may need to set the strength to 1.1-1.25 for the proper effect, due to slight differences in quantization between comfyui and simpletuner.</s> The strength has been fixed with <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI/commit/bb222ceddb232aafafa99cd4dec38b3719c29d7d">https://github.com/comfyanonymous/ComfyUI/commit/bb222ceddb232aafafa99cd4dec38b3719c29d7d</a> Update ComfyUI and the lora will work as it should at 1.0 strenght.This LoRa has been trained gently for highest compatibility with other LoRa's. Trained with a variety of lengths of captions all of which include "microbikini". Dataset included real photos and anime pictures, so it should fit for most peoples use-case.

Files with SHA256: 896baa81a503551a8d762f836b5b021ffce5c8343cb939227994069291548476