A TTS model capable of generating ultra-realistic dialogue in one pass.