Assistant CoreAssistant Core
Web App

Voice Chat

Talk to your AI assistant using Standard or Realtime voice mode.

Overview

Voice Chat lets you have a spoken conversation with your AI assistant directly in the browser. No extra app or hardware needed — just your microphone.

Open Voice Mode

From the chat interface, click the Voice waveform button at the right of the input bar. It appears when the input is empty. The app switches to Voice Chat, connects, requests microphone access when needed, and starts the first listening turn automatically.

Voice Modes

Switch modes from the Standard / Realtime control in the bottom voice bar before a turn starts. See Standard vs Realtime to compare flow, latency, and tool support.

Realtime mode is available on every plan. If the Realtime option is grayed out, no Realtime model is available right now.

Your voice mode preference is saved automatically and remembered across sessions.

Using Voice Chat

  1. Open Voice Chat and allow microphone access.
  2. Speak when the session starts listening.
  3. Select End to stop the current turn, or wait for automatic turn detection in Standard mode.
  4. The AI processes your speech and responds with voice. Text appears when response-text streaming is enabled.

If the assistant admin turns Stream Response Text off, you will still hear audio, but transcripts, response text, or tool-call status text may not appear in the voice interface.

During a conversation

  • Your messages appear as text bubbles (transcribed from your speech)
  • AI responses appear as text and are read aloud simultaneously
  • Tool calls are shown as status messages (e.g., "Calling knowledge_base...")

Type or attach a file

You can type a message in the bottom input. In Standard mode, the + menu can also attach files when file upload is enabled. Realtime mode does not accept file attachments.

Controls

ControlAction
Voice actionStart voice when idle, end an active turn, or cancel startup
MicrophoneMute or unmute audio sent to the assistant
Mode selectorSwitch between Standard and Realtime while idle
ReconnectReconnect after a dropped or failed session

The mode selector is hidden while the AI is listening, processing, or speaking. End the active turn before changing modes.

Return to Text Chat

Select New chat or any saved conversation from the sidebar to return to text chat. Voice messages remain in the same conversation thread.

Requirements

  • A modern browser (Chrome, Firefox, Edge, Safari)
  • Microphone access — your browser will prompt for permission on first use
  • Internet connection

Voice usage consumes credits from your wallet. See Credits for details on voice costs.

On this page