AI art kind of feels like technology that has some really cool potential, but was taken in completely the wrong direction.

Like image generators are just dumb. Describing in text what you want a picture to look like is exactly the wrong approach. We even have sayings about this exact issue! A picture is worth a thousand words. Well even with a thousand word prompt, it’s still incredibly difficult to get the picture you want. This is ignoring all the tracking down of various loras and what not to try to control the output a bit better.

No, the whole image generator concept is just a dumb gimmick of extremely limited value to anyone.

The real value (to me) would be empowering already talented artists with custom loras and LLMs trained on their own art to help speed up repetitive parts of their art creation process.

To enable individual artists to scale up their work into larger logical chunks. Maybe they’ve already mastered noses or eyes or this one specific squiggly line that really comes up quite a lot for some reason. To take someone who’s already talented and give them an even faster and smoother workflow to get to what new and unique idea they’re trying out now faster.

Basically the same way auto complete helped programmers but for artists.

Maybe this already exists, or maybe it’s still being worked on, but to me this seems like the way we should’ve always been headed. Humanity is about building better and faster tools. Not just trying to cope with randomly generated slop.