> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zerotwo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Start a Voice Chat

> How to begin a real-time voice conversation with ZeroTwo — from clicking the mic icon to ending the session.

Starting a voice conversation with ZeroTwo takes just a few seconds. Click the microphone icon, grant microphone access if prompted, and start speaking — ZeroTwo handles detection, transcription, and response automatically.

<Info>
  The first voice connection in a new browser session may take a few seconds to initialize the WebRTC stream. Subsequent connections in the same session start faster.
</Info>

***

## Steps to start a voice chat

<Steps>
  <Step title="Open any chat">
    Navigate to an existing conversation or start a new one at `/c`. Voice is available in all chat contexts.
  </Step>

  <Step title="Click the microphone icon">
    Find the **microphone icon** in the prompt bar at the bottom of the chat window. Click it once to activate voice mode.
  </Step>

  <Step title="Grant microphone permission">
    Your browser will show a permission prompt asking for microphone access. Click **Allow**. This only appears the first time you use voice on zerotwo.ai (or if you previously denied access).

    If you accidentally clicked Block or previously denied access, see [Voice Troubleshooting](/tools/voice/troubleshooting) to reset permissions in your browser settings.
  </Step>

  <Step title="Wait for the listening indicator">
    Once the microphone is live, a **pulsing indicator** appears near the mic icon. This confirms ZeroTwo is actively listening and ready for your voice input.
  </Step>

  <Step title="Speak your message naturally">
    Talk at a normal pace. **Voice Activity Detection (VAD)** automatically detects when you start and stop speaking — no button to hold down. When you finish a sentence or pause, ZeroTwo begins processing.
  </Step>

  <Step title="ZeroTwo responds with voice and text">
    After you finish speaking, ZeroTwo processes your message and responds with a voice reply. The response is also displayed as text in the chat, so you always have a written record.
  </Step>

  <Step title="Continue the conversation">
    Speak again to keep the conversation going. Take turns naturally — ZeroTwo listens after it finishes speaking, but you can interrupt it at any time (see below).
  </Step>

  <Step title="End the voice session">
    Click the **microphone icon** again, or press **Escape**, to end the voice session and return to standard text mode. The transcript from the session remains in the chat history.
  </Step>
</Steps>

***

## Interrupting ZeroTwo

You do not need to wait for ZeroTwo to finish speaking. If you want to redirect the conversation, correct something, or ask a follow-up mid-response:

1. Simply start speaking while ZeroTwo is responding
2. ZeroTwo stops its current response immediately
3. Your new message is processed and ZeroTwo replies

This makes voice conversations feel natural rather than rigidly turn-based.

***

## During a voice chat

Voice mode and text mode are not mutually exclusive. While a voice session is active, you can still:

* **Type a message** in the prompt bar and send it — ZeroTwo responds with voice
* **Paste text** into the prompt bar — useful for sharing a URL, code snippet, or long passage you don't want to dictate
* **Attach files** using the paperclip icon — ZeroTwo reads and responds to them verbally

This is especially useful when you want to share something that is difficult to dictate (like a URL or block of code) without leaving voice mode.

***

## Session indicators

| Indicator                | Meaning                                      |
| ------------------------ | -------------------------------------------- |
| Pulsing mic icon         | ZeroTwo is actively listening for your voice |
| Solid waveform animation | ZeroTwo is currently speaking                |
| Static mic icon          | Voice session is not active                  |

***

## Transcripts

Everything said during a voice session is saved as text in the chat history:

* Your spoken messages (what you said, transcribed via Whisper-1)
* ZeroTwo's responses (what it said, shown as text)

Transcripts are available immediately after each exchange — you do not need to end the session to see them. Use voice for free-form brainstorming and rely on the transcript afterward without worrying about note-taking.

***

## Changing your voice mid-conversation

If you want to switch to a different voice during an active session:

1. End the current voice session (click the mic icon or press Esc)
2. Go to **Settings → Preferences → Voice** and select a new voice
3. Click the mic icon to start a new session — the new voice takes effect immediately

Voice preference changes do not apply mid-session; they take effect when you start the next session.

***

## Ending a voice session

To stop voice mode at any time:

* Click the **microphone icon** in the prompt bar
* Press **Escape** on your keyboard

After ending, you return to standard text chat. The full transcript from the voice session remains in the chat history.

***

## Browser compatibility

Voice chat uses WebRTC, supported in all major modern browsers:

| Browser | Support                      |
| ------- | ---------------------------- |
| Chrome  | Full support                 |
| Edge    | Full support                 |
| Firefox | Full support                 |
| Safari  | Full support (macOS and iOS) |

<Warning>
  Voice chat requires an HTTPS connection. All standard zerotwo.ai pages use HTTPS. If you are accessing ZeroTwo over HTTP for any reason, voice will not be available.
</Warning>