This kind of reminds me of older deep dream style images.
I miss those and the VQGAN images. I wonder if there’s a way to run those original old models and algorithms on newer hardware, faster. I feel like they were more interesting to probe around to see how the model works.
Not that the new stuff can’t be interesting to play with, it’s just that the older ones felt more like toys and less consequential.
I've thought about using this kind of thing for artwork but I don't know how to feel about that. But I'm small and haven't even released anything proper so it's not like I'd pay anyone to do art for something ten people would see anyway.
If I was, I feel like I'd pay an artist for my album covers, no matter the number of fans, for dedication to my art. But I understand musicians often are poor, so I don't know.