Raw LLM Responses

I feel like one of the best things about the current botch of ai "art" generators is just how absolutely clueless the people who make them are about what would make a useful tool for implementing a creative vision. The whole obsession over the idea you just tell it to do something and it just does it makes it seem like it takes more time and effort to get a desired result than actually just doing it. I went to school for commercial photography, so especially seeing the video generator pitches makes me laugh, there's no granularity, no precise control on every aspect of the frame that these things require. But the AI vendors will act like an test commercial they've run (infringing on a major trademark and an actor's likeness in a commercial product, smooth move that won't get you sued) that'd fail a student project is awesome because it was quick. The last thing that drove this in for me recently was an artist complaining on bluesky about a fan who fed their linework piece into AI to color it. And first thing I noticed about the image is that it was a completely different picture, the character was roughly the same size and proportions, but all the details were different. And that really shows that the attempts to compare this to learning from someone else's style is outright false. If the AI were actually learning like a human it would have the ability to fulfill that request without changing the image. It needs to be able to draw from it's library to patch together bits to collage an image together, it can't deal with something outside of that.