LoRA trained on cosplay images of latex Asuka. Might be hit or miss depending on the model used. Each outfit is a standalone model because I was getting poor results trying to train multiple concepts. The dataset was less diverse for the alternate outfits so the face has bad weighting, the white one is the worst in that regard.
Description
eye patch can cause problems even if you don't tag it, so inpainting might be required. Dataset was also limited in size so face/background might be skewed
FAQ
Comments (3)
Nice to see SDXL gettting more coverage. Could you also make latex LoRA for Mari's pink plugsuit? or male plugsuits for Shinji/Kaworu?
Lol, you've trained a whole collection of models based on photos of one cosplayer. I wonder if using ADetailer, is it possible to restore the face completely...
Guy has been AFK for a long while now, the best we can do is re-make the LoRA and hope for the best