I'm curious how this handles bass music where the lower frequencies need to be tightly mixed to sound good. Is that really something that a model can learn?