Voice Providers
Voice providers enable speech-to-speech voice chat for your assistants. When a voice provider is configured, a Talk button appears on the assistant detail page, allowing users to have real-time voice conversations.
How Voice Providers Work
Voice providers are full speech-to-speech agents — they receive audio from the user’s microphone, process it using a realtime AI model, and stream audio back as the response. This is distinct from text-to-speech (TTS) channels, which convert text responses to audio after the fact.
TeamWeb AI currently supports the following voice provider:
| Provider | Models | Description |
|---|---|---|
| OpenAI Realtime | gpt-realtime-2, gpt-realtime-1.5, gpt-realtime-mini | Full speech-to-speech via OpenAI’s Realtime API (GA endpoint) |
Enabling a Voice Provider
Navigate to Voice Providers
Go to Settings → Voice Providers in the sidebar (admin access required).
Configure the Provider
Click Configure on the voice provider you want to enable (e.g. OpenAI Realtime) and paste your API key. Select a default voice model — this is what new assistants will use unless they override it.
Save and Enable
Save the configuration to activate the provider.
Enabling Voice on an Assistant
Once a voice provider is configured deployment-wide, you can enable it per assistant.
- Open the assistant’s edit page.
- In the Voice section, select a provider from the dropdown.
- Select a voice model.
- Select a voice character (e.g. alloy, shimmer, echo).
- Save. The Talk button now appears on the assistant detail page.
To disable voice for an assistant, set the provider dropdown back to None.
Using Voice Chat
With voice enabled on an assistant, click Talk on the assistant detail page to open a voice conversation. See Chat — Voice Chat for details on how the interface works during a voice session.