I did a lot of additional testing on the pony model.
I think I got to a good point of optimizing the prompt however there are some pros and cons.
pros:
prompt: it is very important and it is quite followed in the generation of the image
good anatomy: especially hands and feet, not perfect but definitely better than SD 1.5 models
NSFW: definitely good
cons:
faces/eyes: some are a bit weird, I need to see if I can fix it via prompt optimization/refinement
generation time: definitely longer than SD 1.5 models, we are talking about 4-5 times slower. i need to figure out if i can somehow speed it up.
upscale: the upscale method I normally use on SD 1.5 doesn't work on the pony, I've done a lot of experimenting and I think I've come up with a decent compromise but there is definitely room for improvement.
also tell me what you think, whether I should continue to invest time on it or give up and continue with the usual SD 1.5 models
AnythingForYouMyDear
2024-06-11 00:08:08 +0000 UTCai-freegirls
2024-06-10 06:50:29 +0000 UTCAnythingForYouMyDear
2024-06-09 21:11:34 +0000 UTCJohn Doe
2024-06-08 18:38:21 +0000 UTCRick
2024-06-08 13:11:34 +0000 UTCSpinal121
2024-06-08 11:49:37 +0000 UTC