I discovered something interesting about how the image generator works. After a bunch of testing, I started noticing that it absorbs way more from the written prompt itself than from analyzing the image it created. It kind of tries to infer your intentions directly from the prompt. So when you use a huge, super detailed prompt to generate an image, it tends to be less restrictive or less “aware” of certain things, which can lead to images your usual prompt wouldn’t be able to generate. I even managed to get a full topless image myself just by asking for artwork of Aphrodite.
Now I’m curious if anyone else has noticed this too, or wants to take the testing further
[link] [comments]