What is the best TTS for realtime voice agents?
Connect Teen V1 to LiveKit, Vapi, Pipecat, or your own realtime stack through standard HTTP and WebSocket speech interfaces.
What does a live voice agent need from TTS?
Realtime-friendly output
Use PCM and streaming delivery when the agent needs to begin speaking before the full response is ready.
Performance follows context
Direct a reply as calm, playful, warm, precise, or energetic without changing the agent architecture.
Multilingual turns
Serve global and code-mixed conversations with one model instead of routing every language to a different vendor.
How do I add Teen V1 to a voice agent?
Receive the agent response
Pass the final text from your model or orchestration layer into the TeenageWorks speech request.
Stream audio into the transport
Forward PCM or streamed audio chunks to the room, call, or media pipeline.
Cancel when the user interrupts
Close the active stream on barge-in and begin the next response without replaying delivered bytes.
What else should I know?
Does this work with LiveKit, Vapi, and Pipecat?
Yes. TeenageWorks can be connected through their provider hooks or a custom HTTP or WebSocket adapter.
Which format should a realtime agent use?
PCM is the most direct choice for realtime audio pipelines. MP3 is convenient for web playback and WAV for complete files.
How should interruption be handled?
Cancel the active stream when the user interrupts. Do not automatically retry a partially delivered stream because that can duplicate speech.