Realtime Voice
Realtime Voice lets you have a live spoken conversation with your WednesdayAI assistant — streaming two-way audio with near-instant responses, live transcripts, and the ability to interrupt the assistant mid-sentence.What you need
A voice provider must be configured by your WednesdayAI administrator before realtime voice is available. If no provider is configured, the feature will not appear or will return an “unavailable” message. Ask your admin to install and configure a realtime voice provider plugin.How a session works
- Start — your client app sends a session request; the Gateway connects to the voice provider.
- Speak — your audio streams to the Gateway in real time.
- Listen — the assistant’s audio streams back to you with a live transcript.
- Interrupt (barge-in) — start speaking at any time to interrupt the assistant’s response.
- End — tap Stop or disconnect; the session closes gracefully.
During a session
Session limits
- 30 minutes maximum per session. The session ends automatically with a “completed” notice.
- 2 concurrent sessions per connection — starting a third requires closing one first.
Interrupting the assistant
You can speak while the assistant is responding — this signals a barge-in and stops playback. The assistant processes your new speech from that point. Some providers support barge-in detection automatically; others require you to tap a button in the client UI.Assistant actions during voice sessions
If your assistant is configured with agent access, it can perform actions while you speak — looking up information, checking memory, reading files, or running tools — and report the result back as a spoken response. You may hear a brief acknowledgement (“let me check on that”) while the action runs.If something goes wrong
Contact your WednesdayAI administrator if sessions consistently fail. Admin configuration
instructions are in Realtime Voice — Gateway setup.