Trained for 10 epochs at network dim 128 on the animefull-final model, resized to a network dim of 64. This LoRA works with pretty much any derivative merge of the animefull model (eg. AnythingV3, AOM2, Chilloutmix, etc).

The outfit and character were trained so that the model could recognize them as separate concepts. Meaning you can put the outfit on other characters or give Mai an entirely different outfit.