>>55019409>novelWhoops, meant DALLE. But let me give a quick overview
because i wrote it up before i caught my mistake.
You can just give it keywords, but you'll hit walls. The way I think it works is
>ai checks for banned words>ai checks for bad implications (sexual or otherwise)>ai spellchecks>ai checks visually if it thinks it's bad contentTo get past the 2nd part, you have to trick it to thinking things are safe. You can also give it implications it thinks are safe on their own but gives a breakthrough together. Like 'suds, soap, bubbles, foam, steamed' might get you a bath scene when bath words are banned. Describing things as soft or gentle (like lighting or scene) helps.
3rd check is like a cheat. Misspelling gets you further. Relaxing? Bad. Relaxin? Perfectly okay, here's a girl lounging with butt up. Lots of testing to make work.
But 4th is just the annoying wall. If it thinks you have sexy stuff, out it goes. Hard colors and shadows and foreground objects help you get away with things.
It's not going to have its full potential realized until a perv rips it away from corporate ownership.
