Voice Input
Speak instead of typing. The transcript lands in the message box as ordinary text, so you can read it, fix it, and send it when you're ready.
Where the button is
The mic sits inside the composer pill, between the text box and the Send button. The + attach button is on the left of the same pill.
It's there on desktop and on phones, and in every chassis β all 12 share the same composer.
The mic is unavailable while a reply is streaming. Wait for the reply to finish, or stop it first.
Next to the mic is a small gear that opens Voice settings (see below).
Two ways your speech becomes text
Nyquest prefers your browser's own built-in speech recognition. If the browser doesn't have it, Nyquest transcribes the recording for you instead. You don't choose β the composer picks whichever is available.
| Browser's built-in recognition | Nyquest transcribes it for you | |
|---|---|---|
| Browsers | Chrome, Edge, Safari, and most Chromium-based browsers | Everything else with microphone recording, Firefox included |
| What you see | Words appear live in the status strip as you talk | A Transcribing⦠status, then the whole passage at once |
| Speed | Immediate | A short wait after you stop |
If a browser has neither, the mic reports that voice input isn't supported and you'll need to type β or switch to Chrome, Edge, or Safari.
While it's listening
A status strip appears above the composer with a pulsing dot, a timer, a level meter, and β on the built-in path β the words being recognized so far.
| To do this | Do that |
|---|---|
| Stop and keep what you said | Click the mic again (it shows a square while listening) |
| Cancel and discard it | Click the X in the status strip |
| Dismiss an error message | Click its X, or press Escape |
Voice input also stops on its own after five minutes, and whenever you switch away from the tab. The browser's recording indicator never outlives the session.
The transcript goes to the composer, not to the model
Nothing is sent when you stop talking. The text is appended to whatever is already in the box β spacing and capitalization handled for you β and it behaves exactly like text you typed. Edit it, add to it, then send.
If you'd rather not press Send, turn on Automatically send after speaking. You then get a short Sending in⦠countdown with a Cancel button before the message goes. Very short fragments and filler-only sounds ("um", "uh") are never auto-sent.
Voice settings
Open them with the gear beside the mic.
| Setting | Notes |
|---|---|
| Automatically send after speaking | Off by default |
| Spoken language | English (US/UK), EspaΓ±ol, FranΓ§ais, Deutsch, Italiano, PortuguΓͺs, ζ₯ζ¬θͺ, δΈζ. Defaults to your browser's language |
| Microphone | Only appears once you've granted mic permission at least once β before that, the browser's default device is used |
These settings are stored in the browser you set them in. They don't follow you to another device or another browser.
Microphone permission
The first time you click the mic, the browser asks for permission and the strip reads Waiting for microphone permissionβ¦. Allow it once and the browser remembers for the site.
If you deny it β or denied it in the past β voice input fails immediately with Microphone access is blocked. Clicking the mic again just re-fails, because the block is at the browser or OS level. Re-grant it there:
- Chrome / Edge β click the icon at the left of the address bar, Site settings β Microphone β Allow, then reload the page.
- Safari (macOS) β Safari β Settings for This Website β Microphone β Allow.
- Safari (iOS) β Settings β Safari β Microphone.
- Firefox β click the padlock in the address bar and clear the blocked microphone permission, then reload.
If the browser never even asks, check the operating system: macOS System Settings β Privacy & Security β Microphone, or Windows Settings β Privacy β Microphone, and make sure your browser is allowed there.
One more thing worth knowing: only the app origin is permitted to use your microphone. Our other hosts deny it outright, so use the app itself rather than any embedded or proxied copy of it.
What happens to the audio
- The microphone is open only while voice input is active. Stopping, canceling, hiding the tab, or leaving the page releases it.
- On the built-in path, recognition is handled by your browser using whatever speech service it ships with. The audio never reaches Nyquest.
- On the fallback path, the recording is used for that one transcription request and nothing else. It's held in memory for the length of the request, never written to disk and never logged. The transcript isn't logged either β only metadata such as size, duration and whether it succeeded.
- There is no voice history. No audio file is kept on your account, and recordings can't be replayed later.
When something goes wrong
| Message | What it means |
|---|---|
| Microphone access is blocked | Permission denied. Re-grant it as above, then reload |
| No microphone detected, or it is in use by another app | Check the device is plugged in and close anything else holding it (calls, meetings) |
| Voice input stopped: network problem | Connection dropped mid-session. Try again |
| We could not transcribe that recording | The transcription attempt failed. Try again, ideally with a shorter passage |
| Voice input stopped. Try again | Built-in recognition ended the session by itself. Nyquest restarts it a few times automatically; this appears once that runs out |
| Voice input is not supported by this browser | No built-in recognition and no recording support. Type instead, or use Chrome, Edge, or Safari |
Quiet input is the usual cause of a session that ends early β the level meter tells you whether the mic is actually hearing you.
Where to next
- Streaming β how replies come back
- Slash Commands β the other composer shortcuts
- Conversations and History β where your messages end up