A style based on NSFW furry art by Cervina. Basic prompt is "furry female", then you can just describe anything that is not being generated consistently (huge breasts, wide hips, tail, hair/fur color, tattoos etc., full tag list available through A1111 LoRA menu) Tested this mainly in non-specialized CardosAnime checkpoint and except trouble with generating tail on images with view from behind, it does surprisingly well. Short test in YiffyMix as well, it's much easier there if you are not asking for too much flexibility between checkpoints.
Description
FAQ
Comments (2)
Was wondering what model you trained with to make the LoRA?
I saw you mention CardosAnime or YiffyMix for testing inference, was it one of those?
No, those were really used just for the LoRA testing and creating sample images visible here - I'm training everything with the base SD 1.5 :)