Trained on images of women trying to hide their assets from prying eyes.
Strongly recommend hires fix.
I'd recommend starting with the trigger caught naked. General non-trained prompting can help solicit different draws, like hunched over or covering.
caught by surprise can help improve the likelihood of the associated expression, especially with adetailer.
This version works the best with high-context subjects (like LoRA's), so if you are doing a generic person, it might help to add more details about them.
Shouldn't need meaningful negative prompting, but disfigured can help reduce the occasional 3-arm image.
Description
FAQ
Comments (6)
Hi, what model are you using in your realistic example images?
ICBINP Relapse for most. Sometimes Realistic Vision 4.0. Training done against base model
It seems to always make them with their arms folded under the breasts. No covering, and not even attempting to cover the groin.
What model are you using? A bunch of the shared generations capture that pose pretty commonly.
@thegipper tried a couple. yiffymix, darksushi, epicphotogasm. it can do these poses but it's reluctant.
@renakunisaki same






