I like doing my own composition, but AI being able to just do sound design from a prompt ("hey can you make a synth that sounds like the lead from XYZ song but buzzier, add parameters to control ABC") would be revolutionary.

> This being text based makes it really easy to have AI generate the music.

It's already really easy for AI to generate music without using text as an intermediate format.

Before long we can get agents to listen to the music, too. Then based on our prior data can tell us how much we liked it.