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

# Voice conversations

> Talk to your AI using voice on supported channels.

Voice conversations let you speak naturally instead of typing. The AI listens in a voice channel, turns your speech into text, thinks it over, and answers out loud.

## Supported channels

Voice is available on channels where the administrator has enabled it:

* **Discord** — voice and stage channels in servers where the AI is set up
* Other setups where your administrator has connected a voice service

> [!NOTE]
>
> Your administrator must enable voice before these options appear. If you don't see voice options, contact them — setup instructions live on the [administrator voice page](https://docs.wednesdayai.dev/next/admin/gateway/realtime-voice.md).

## Starting a voice session

A voice session is the time between the AI joining your voice channel and leaving it. On Discord, you control it with the `/vc` commands:

**1. Join a voice channel**

Click a voice channel in your server, the same way you would for a voice call with friends.

**2. Invite the AI**

Type `/vc join` and pick the voice channel from the options. The AI joins and confirms with a message like "Joined `<channel>`" that only you can see.

**3. Start talking**

Just speak. There is no push-to-talk button. When you pause, the AI treats your pause as the end of your turn and answers.

Other useful commands:

* `/vc status` — shows which voice channels the AI is currently in
* `/vc leave` — the AI leaves the voice channel

Voice commands only work in servers, not in direct messages.

## What to expect

* **Speak, pause, get an answer.** When you stop talking, about a second of silence marks the end of your turn. Then the AI answers out loud in the channel.
* **You can interrupt.** If you start speaking while the AI is talking, it stops mid-sentence and listens to you. You never have to wait for it to finish.
* **Short sounds are ignored.** Coughs, clicks, and other noises under half a second are skipped, so accidental sounds don't become questions.
* **Answers are spoken, not typed.** The reply plays in the voice channel rather than appearing as a text message.

> [!TIP]
>
> Interrupting is normal and expected. If the AI is going down the wrong path, just start talking over it and redirect.

## Tool calls during voice

When your request needs the AI to look something up, check a file, or run a task, it uses the same tools it would use for a typed message. From your side this looks like:

* A short pause while it works — the more complex the request, the longer the pause
* Sometimes a quick spoken note that it's checking on something
* The real answer spoken aloud once the work is done

Nothing is posted to the text channel during a voice session; everything you hear is spoken.

## Privacy

Your speech is turned into text and handled like a regular message in your conversation history. Voice works best in quiet channels — background conversation may be picked up as speech.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.