# Use your voice

Dictate a message with the microphone in the composer, have an answer read aloud, and turn on auto-play so every answer is spoken as it arrives.

You can speak a message instead of typing it, and have any answer read to you. Both run through speech services your administrator has configured on the platform.

**Who can use it:** the **Member** role. Your administrator can turn voice features off for the whole workspace; then the microphone does not appear in the composer, and **Read aloud** falls back to your browser's own voice.

Video: Use your voice (2:27), https://katonicdocsmedia.blob.core.windows.net/docs/voice/1ebf90c78d68/index.html

## Prerequisites

- A microphone, and permission for the site to use it. Your browser asks the first time you select **Voice input**.
- Sound on your device to hear answers.

## Dictate a message

![The composer at the bottom of a conversation: the message box with its placeholder, Attach and the agent chip on the left, and Voice input (the microphone) on the right, with the end of the answer above it.](https://docs.katonic.ai/img/user-guide/work-room/voice/composer-voice-input.png)
*Voice input sits at the right of the composer while it is empty. Typing or attaching a file replaces it with Send.*

**Voice input** (the microphone) sits at the right of the composer while the composer is empty. As soon as you type or attach a file it is replaced by **Send**, and while an answer is streaming by **Stop generating**.

  
**Step 1. Start recording**

Select **Voice input**. The first time, allow the browser to use your microphone. A ring around the button moves with your voice, and a timer above it counts the recording.

  
**Step 2. Speak**

Say your message. Recording stops by itself after about two seconds of silence, or after two minutes. Select the button again (**Stop recording**) to stop sooner. **Cancel recording** (the small cross beside the button) throws the recording away.

  
**Step 3. Wait for the transcript**

The button shows **Transcribing…** while the recording is turned into text. The text is then sent as your message straight away, appears in the thread, and the agent answers it as it would a typed message. A recording too short to contain speech is dropped and nothing is sent.

If the browser refused the microphone or the transcription failed, the button shows a warning icon named **Voice input error, try again** for a couple of seconds and then returns to normal. Select it to try again.

## Listen to an answer

![The icon buttons under an answer: Good response, Bad response, Copy, Read aloud and Regenerate response, with Read aloud under the pointer; above them a follow-up chip and the end of an approval card.](https://docs.katonic.ai/img/user-guide/work-room/voice/answer-read-aloud.png)
*Read aloud (the speaker) is the fourth icon under every answer. While an answer plays it becomes Stop reading.*

  
**Step 1. Read aloud**

Under any answer, select **Read aloud** (the speaker). The button spins while the audio is fetched, then becomes **Stop reading** with a progress ring while the answer plays.

  
**Step 2. Stop**

Select **Stop reading**, press Escape, or open another conversation. Starting to read another answer stops the one playing.

  
**Step 3. Hear every answer automatically**

Turn on **Auto-play responses** on the **Voice** tab of Preferences. Each finished answer is then read as soon as it arrives. The **Voice** and **Speed** you pick there apply to auto-play and to **Read aloud** alike; see [Preferences](/user-guide/work-room/preferences).

What you hear is the text of the answer with its formatting removed. An answer that ended in an error is not read automatically, the agent's reasoning is never read, and auto-play cuts a very long answer at about 4,000 characters, at the end of a sentence where it can.

## Verify

Your spoken message appears in the thread as text, and the agent replies to it. While an answer plays, the speaker button under it reads **Stop reading**.

## Troubleshooting

- **There is no microphone in the composer.** The composer holds text or an attachment (send it or clear it), an answer is still streaming, or your administrator has turned voice off.
- **Nothing was sent after I spoke.** The recording was too short, silence ended it before you spoke, or the transcript came back empty. Try again, closer to the microphone.
- **The button shows a warning.** The browser refused the microphone or the transcription failed. Allow the microphone for this site in your browser's settings and select the button again.
- **Read aloud sounds like my browser, not the voice I chose.** The platform's speech service did not return audio, so the browser's own voice was used. Try again later, or tell your administrator if it keeps happening.
- **Playback does not stop.** Press Escape or select **Stop reading**.

## Related

- [Preferences](/user-guide/work-room/preferences)
- [Home and chat](/user-guide/work-room/home-and-chat)
- [Keyboard shortcuts](/user-guide/about/keyboard-shortcuts)
- [Cards and reasoning in answers](/user-guide/work-room/cards-and-reasoning)

Source: https://docs.katonic.ai/user-guide/work-room/voice
