NSFW!!
LoRA trained on ~20 cherrypicked images, intended to improve futanari style images. It does make a difference but let me know if this LoRA is any good or not.
For prompts, copy some from the showcase images.
Description
First version.
This will work on any NAI-based model (animefull, Anything, AOM2 etc)
Some training parameters for those interested:
DIM size: 32
warmup: nothing
scheduler: cosine
FAQ
Comments (1)
This works really well actually, which has me curious on how i could learn to do this with other artist, im trying my hand a "Taigararts", but have no clue how you got this to even work haha