Made something cool with this LoRA? Share it in the gallery below to show you like it!
What is the LoRA model for?
Replicating Takahirosi's style, including lighting details (caustics, diffuse lighting, reflections, etc), anatomy (thick muzzles and big muscular bodies along with very prominent glans on penises) and vibrant colors among other things. Capable of making rubber/latex suits of different colors as well, adding appropiate tags. Excels at furry men, but should be able to make female furries as well as male and female humans too with some work.
Usage and recommended settings:
Works well on ComfyUI. Will work well on Auto1111 too, however you will have to tweak settings to make it work as fully intended as Auto1111's prompt parser and CFG scaling works differently than in ComfyUI.
For SD1.5 Models:
Strength for Model and CLIP: 0.77 (Range of 0.64-0.92)
Keyword is ta-hiro, with recommended weight of 1.2 [meaning (ta-hiro:1.2) in-prompt] works well at the very beggining or end of the prompt.
SD 1.5 only. VPred and ZSNR trained models seem to be compatible too. A model finetuned on furries such as Indigo Furry Mix v12.0 Hybrid, EasyFluff, Fluffusion or FluffyRock will work best.
For IndigoXL and PonyXL:
Description
Version for SD 1.5
FAQ
Comments (4)
Great model, with a unique depiction of (rubber body).
Thanks for the support! Will train an XL version soon.
@blidbear I'm surprised to hear that. I used this model since I came to this website, but I didn't use it because there was no higher resolution in online generation. I use indigo's checkpoint, which is not as high as Pony resolution. So can the new version be used on Pony?
@zy02 I will train two versions: One for CompassMix (Should also be compatible with SeaArtFurryXL and BananastrikeXL) and another one for IndigoXL v3 (I think it may also be kind of compatible with Pony, since it is in the mix). I will explicitly train on Pony once the version of it trained with corrected training data comes out, as the current one has issues (the "score_XX" and "BREAK" stuff, for example).
Also "soon" is a bit of a lie. There is 1 hr left for training at the time I'm posting this comment. I will do some testing and then publish it in a few hours.