>>537567815
you can strait up ask for certain values it assigns to images, and there is a factor of "lewdness" where is the resulting image would bypass a certain rating it will prevent the generation or will actively censor the generation to remain in the limits. it does it at the start with the insert image, and it repeats this check for every requested detail. AI chat models are retarded where you can ask them info about their code backend and they will give it to you if you format the request good enough to bypass the manual blocks put into place. it's why the lewder/more detailed you want to abuse this shit with, the more limited the image has to be so the vocal point for the lewdness can be focused on. a weaker style will allow more lewdness, and an image with a bust up focus will have more ways to focus on it so the "lewd" value can be maximized with having a better generation. it's the same thing with most AI models in some shape or form since AI designers are all hacks who are scamming the US government making them believe shit shit is worth trillions of dollars.

anyways here is a noot in a soot.