My favorite AI songs are those that have been created with human intervention and creativity, not just button mashing
Early Udio generated songs when run through a stem splitter were horrible to listen to and you could hear how it was just generating the frequencies with none of the texture—like a simulation of the frequencies an instrument would produce.
At the very least they should be re-recorded with real instruments and vocals or using the existing digital tools. Slop is one word; I like to use mush, as that is what it sounds like when you really listen closely.
The sound is amazing if you’re doing it right. I produce music using the old fashioned way, with a DAW, and have done for decades, and there are some songs that I’ve made with AI that I go to again and again because they sound great.
Basically, what we’re seeing is the personalization of media.