Power LoRA becase I wasnt satisfied with the others. Loads of NSFW refs, because cmon, who are we kidding. Trained on a fair few dymanic images, but also some of the toilet based ones, so should be very easy to get toilet based images.

Trained on Anything v4.0.

INPAINT EYES - Can fairly accurately capture the character, but the big thing that I wanted was the eyes. So while sometimes the eyes may be a little strange, all you need to do is send it to inpaint, and (with the LoRA) use 'cross-shaped pupils, symbol-shaped pupils'. Paint the entire eyes, not just the pupil, use only masked, reduce the quality down to 512x512, and a denoise of around 0.5.

The teeth are also sometimes hard to get good. You can also do the same Eye inpaint steps for the mouth if you wish.


Weighting should be good around 0.7-0.8.

PLEASE NOTE THAT A LOT OF THE REF IMAGES ARE INPAINTED, THUS MISSING THE ORIGINAL METADATA

Enjoy