The thing that bums me out about AI image generation is how quickly all the models converged on the same hyper realistic and overly saturated aesthetic. I suspect they’re all using more synthetic training data which lacks the subtle imperfections of reality.