LoRA trained on over 200 curated and extensively tagged Ankha images.

Trained on top of the meinahentai_v4 model, which has an anime look

Both the images and base model used for training are quite lewd, so expect that in your output

Seems to work best with alphas set at 0.5 +- 0.1

Planning on tweaking the fitting in future versions, but you can still get some pretty good results as it is now