What you want is the model to be so advanced that it would probably be capable of just generating music. And actually if all you want is just similarly sounding music streaming into your ears then automatic generation is the natural continuation after automated recommendations.