Voice
Voice
How people talk to your assistant by voice, the two voice modes, and who configures what.
Voice lets people speak to your assistant and hear it answer, instead of typing. Voice conversations are saved in the same history as text chats.
Where voice works
| Where | How people use it |
|---|---|
| Web App | The Voice chat button in the message box. See Voice chat |
| Devices | ESP32 devices talk to the assistant by voice. See Devices |
| Your own app | The Developer API voice WebSocket. See Voice API |
Two voice modes
| Mode | How it works | Choose it for |
|---|---|---|
| Standard | Speech is transcribed, the chat model answers, and the answer is read aloud | Accurate answers, files, the greeting, lower cost |
| Realtime | One voice model listens and speaks directly | The most natural, lowest-latency conversation |
See Standard vs Realtime for the full comparison.
Who configures what
| Who | What | Where |
|---|---|---|
| Assistant owner | Turn voice on, default speech models and voices, recordings, response text, greeting | Enable voice |
| Each signed-in user | Their own models, voices, speech speed, and spoken language | Model and voice settings |
Voice uses credits for speech recognition, speech output, and Realtime minutes. See Credits.
Last updated on