A basic image segmentation model (Yolo-v8) trained on around 200 images. Has some benefits over the kind of model that uses bounding boxes (The squares you see during generation) since it only inpaints the target area (belly). Trained primarily on realistic images, but can do furry / anime too. Sometimes it will recognize round things as bellies, but otherwise works well enough.

Instructions:

Unzip the archive and place the .pt in the /models/adetailer directory of your install

If it isn't working go to your stable diffusion folder and try:

venv\Scripts\activate 

pip install --upgrade ultralytics  

Note this is for windows, if you are on linux I trust you already fixed this