CompVis / stable-diffusion

A latent text-to-image diffusion model
https://ommer-lab.com/research/latent-diffusion-models/
Other
67.5k stars 10.07k forks source link

lack of support of furry species name #243

Open tails101 opened 2 years ago

tails101 commented 2 years ago

i tried to use stable diffusion to generate some image of some popular furry species but it doesn't seems to understand them. something like avali , protogen, wickerbeast, rexouium ... and just being descriptive with some of those specie can be hard. (results sometime looking like a pile of random specie part or random robots for the protogens)

i suggest more training in furaffinity, deviant art or e621 for furries species support. grid_00009_3188344855_anthropomorphic_fox_wearing_a_motorbike_helmet (descriptive version of a protogen.)

KostyaAtarik commented 2 years ago

Try to condition generations not by text but by the images of such species

7ih commented 2 years ago

bro no way ur tryna make furry art with this ☠☠☠☠

tails101 commented 2 years ago

bro no way ur tryna make furry art with this ☠☠☠☠

bruh... youll be surprised aesthetic_portrait_commission_of_an_albino_male_fu_56238355

a_photograph_of_a_anthropomorphic_black_wolf_weari_8ed7f7c7