6
Just realized feeding my AI image generator a rough sketch first gets way better results than a long text prompt.
I was trying to make a picture of a robot fixing a car for 2 hours with text, but after I drew a simple stick figure sketch and uploaded it, the AI got it right on the first try, so is the future of prompting more visual than verbal?
3 comments
Log in to join the discussion
Log In3 Comments
garcia.cameron7d ago
My buddy was trying to get an image of a specific fantasy castle for a game he's making, and he spent a whole night typing every word combo he could think of. It kept giving him weird modern buildings with a single tower. He finally just drew a quick, terrible box with some triangles on top in MS Paint and fed it that. The AI spat out almost exactly what was in his head. It really does seem like these things are starting to understand pictures better than our messy language.
3
henryr458d ago
That's wild, makes total sense though. You think it's just better at understanding spatial stuff from a picture than from our weird word descriptions?
1