ribcaca said:
did anyone have any luck using stable diffusion to make hypno lewds?
Yeah, there's 700+ ai generated pics on the hub (ai art tag)
Others have mentioned that there are good LORAs to use:
civitai.com/?query=hypnosisI've gotten mixed results with them. I got better results from manips though.
GuileGaze said:
Ezrael said:
Did anyone try training a LoRA from hypnohub's images?
I made an one extremely quickly without any manual tagging (just used WD Tagger), but didn't get great results. Especially when dealing with eyes/expressions, you'll probably need to manually tag those in the images or the autotagger will miss them. Not to mention that you'll probably also want a lot of close-ups to properly train those small details.
I might try to go back to it though, especially because none of the models I've found can create a good expressionless/mindless look.
Same! I find that the models are best at creating smiling people. Your AI stuff is really good, by the way! Big fan.