Voice stack

The cascaded row

Diagram SVGDiagram PNGDiagram HTMLTrack map HTML

LiveKit's pipeline page names a cascaded pipeline as speech-to-text, then a language model, then text-to-speech. A realtime model, on that page, consumes and produces speech in one model. Pipecat's overview lists a transport, speech recognition, a language model, speech synthesis, and a transport again.

Sort the cascaded order

Put the first voice room's flow back in order. A speech-to-speech model is a different shape. It does not sit inside this row.

• Speak Up Down
• Decide Up Down
• Transcribe Up Down
• Hear audio Up Down

Match the words

Rooms

1. Speech in, a model, speech out
2. Turn-taking and barge-in
3. Latency budgets you label
4. Telephony and WebRTC
5. LiveKit Agents and Pipecat
6. A hosted voice agent, from the xAI docs

Latency sentences stay attached to the page that wrote them. This lab states no service-level agreement and no price.