Voice mode
Let users talk to your AI guide out loud, with the cursor pointing at what it's describing.
Voice mode turns the guide into a spoken conversation. Instead of typing a question, a user talks — and the guide talks back, while the cursor points at the parts of the page it's describing.
Starting a conversation
Users start voice mode two ways:
- Pressing Cmd/Ctrl + /
- Clicking the mic button in the launcher
Either one opens a real-time spoken conversation, powered by a realtime speech model. There's no transcribe-then-respond delay — the user talks, the guide listens and replies, back and forth, like a call.
Mic permission
The browser asks for microphone permission on first use. This prompt is triggered directly by the user's click or keypress — SupaMaus never requests the mic on page load or in the background.
What the user sees
While the conversation is happening:
- The cursor points. As the guide talks about something on the page, the cursor moves to it — the same pointing behavior as a typed answer, just synced to speech instead of text.
- A waveform reacts. An animated waveform sits next to the cursor and reflects whoever is currently speaking — the user or the guide — so it's always clear who has the floor.
- Captions render live. The guide's spoken words are captioned in real time, under the waveform, as it talks.
Captions are on by default. Site owners who don't want them can turn them off with the data-voice-captions attribute on the embed script:
<script
src="https://www.supamaus.com/embed.js"
data-site-id="yourdomain.com"
data-voice-captions="false"
async
></script>Interrupting
Users don't have to wait for the guide to finish talking. Barge-in is supported — start talking mid-sentence and the guide stops and listens.
To leave voice mode entirely, press Esc.
What's stored
Audio is streamed only for the duration of the conversation. SupaMaus does not store voice recordings.