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.

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.
The strength has been fixed with https://github.com/comfyanonymous/ComfyUI/commit/bb222ceddb232aafafa99cd4dec38b3719c29d7d 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.