Does it make any difference?
#1
by BingoBird - opened
From reddit discussions, do these make any difference?
I suspect I get more mutants with these.
No idea. You could test it yourself with my space, including this uncensored-text-encoder: https://huggingface.co/spaces/WeReCooking/Flux-Klein-4B-CPU vs https://huggingface.co/spaces/black-forest-labs/FLUX.2-klein-4B
Nekochu changed discussion status to closed
I don't have inference speed to do systemic tests, but i seem to notice more mutants at some dimensions WxH than others.
The main driver of mutants seems to be combining multiple pose instructions for multiple actors. I expect we'll someday see text encoders that output 3d scene information including poser (controlnet?) wireframes that get interpreted.
Maybe I should call up BFL... hm..