A lora trined on images of "bad taxidermy". It turned out fun as expected :D V1: trained on just 30 images, it's fun :D V2: trained on a larger dataset (about 100 images). I'm not sure it's significantly better than V1 but in theory it should be more flexible

Bad Taxidermy for Flux

Trained on a bigger dataset. i made various tries. i am not sure it is a net improvement from V1, but it's funny. Not always it makes the animal you want, but hey! with bad taxidermy in play, it's not always clear what are you looking at!

Files with SHA256: df4c89e03ba588aef7c5d31fa740c9fdd4781c3b022b818f62761da860da6c28