A bit dark artstyle focused on cute 4-legged creatures, can do anthro too.Trained on 17(and 18 variations of them) pokemon artworks. No ponies or other stuff. NovaFurryXL v7B was used as base model.Trigger is "dst-mk2".Use "shaded" or "colored sketch"(or both) for slight style variation. "shaded" provides more finished look with smoother lines.If model breaks anatomy or in some other way, consider lowering strength to 0.6-0.9.This model is extremely overfit, but I think its overfit in a good way. It likes to exaggerate things, amplifying some of checkpoint data, despite not having such things in the dataset. Lowering strength helps to tame it.Put "feelers" in negative if you are getting Sylveon's ribbons on other characters. It helps to reduce ribbon clutter when prompting Sylveon too.Surprisingly, the backgrounds it generates look pretty nice (in my opinion), despite not being trained in that direction at all, there wasn't even that many of them in the dataset. They look drastically different when you compare outputs with and without the LoRA using the same seed.

Files with SHA256: 88b019561dd131249e5922f9c1fabd5c1859e80b2721e5a8e7c373c1a1e6579f