google's omni model + podcast ugc is actually blowing my mind this wasn't edited together i wrote one conversation, pressed generate, and got back a complete talking podcast in a single output after dozens of tests i finally landed on the workflow that makes ai hosts sound