v1 - original singing model v2 - conversational model can’t sing (no pitch)