It seems kind of limiting to have the model generate 2:11, maybe give us the option to give the model a "time frame" slider, and the model can generate a song that ends within that certain time frame? For example, generating a song that ends between 3:00 and 3:30.