A 2 billion parameter text-to-speech model by Nari Labs, optimized for conversational and dialogue-based speech generation.