I couldn't find a good SDXL tan line LoRA so I tried making one. It seems to work fine at 0.6-1.0 strength.
Trigger word: "tan lines"
Sometimes it struggles with whether the lines should be clothes or not, e.g. you might get a strap instead of a tan line from a strap.
Trained on 406 images (mix of SFW and NSFW), 5 repeats, 8 epochs with a batch size of 2, constant learning rate of 4e-7 and 500 warmup steps. Resized from 256 dim to 64 dim. No watermarks.
