Skip to main content

Realtime voice overview

Realtime voice lets you speak with WednesdayAI and hear replies as they are produced. You can follow live captions, interrupt a spoken reply, and continue without waiting for the whole reply to finish. For browser instructions, including microphone access and troubleshooting, use Talk realtime voice.
Using the /v2/client web client instead of the Control UI? Its Dictate and Voice mode message-box controls are a separate feature with their own instructions — see Using the web client.

What you need

The person who manages WednesdayAI must turn on realtime voice and connect a supported voice service. The current built-in browser service is OpenAI Realtime. Until it is ready, the Talk tab stays hidden or your app reports that voice is unavailable.

What happens in a conversation

  1. You start voice mode and allow microphone access if your app asks.
  2. You speak and see temporary captions while your words are recognized.
  3. WednesdayAI replies with audio and captions. Final captions replace temporary wording.
  4. You can speak during a reply to stop queued playback and begin the next turn.
  5. You stop voice mode when finished; the microphone is released.
Some apps can also let WednesdayAI look up information or take an allowed action during a voice conversation. You hear the final answer after a short pause. What it may do depends on choices made by the person who manages your WednesdayAI installation.

Limits and reconnects

  • A voice conversation ends automatically after 30 minutes.
  • One connection can have no more than two active voice conversations.
  • If the browser Control UI reconnects, it starts a fresh voice conversation while keeping captions already shown. Select Stop during the disconnection to prevent an automatic restart.

If something goes wrong

Administration instructions are in Realtime Voice gateway setup.