Inworld AI launches Realtime TTS-2 for live conversations

Inworld AI launched Realtime TTS-2, a rebuilt text-to-speech model for live dialogue, not narration.

The model uses full conversation context, supports natural-language prompting and inline audio controls, and keeps one voice identity across 100+ languages.

It is available through the Inworld API and demo integrations with platforms like Layercode, LiveKit, NLX, Pipecat, Vapi, and Voximplant.

The launch targets production voice agents and interactive apps that need low-latency, emotionally aware, multilingual speech.