Your voice,
transcribed.
Nothing leaves
your device.
Scribe turns speech into text entirely on your phone, Mac, or PC. No account, no servers, no listening in. Dictate, record meetings, label speakers, and translate. All offline.
// also on macOS & Windows · iOS soon · no telemetry, ever
Let's ship the beta this week. Everything runs locally, so we don't owe anyone a privacy review.
Agreed. And the recording kept going with the screen off. The background capture works.
Translate it to Hindi for the notes and we're done.
Everything Wispr Flow and Superwhisper do, without sending your voice anywhere.
Live dictation
Talk and watch text appear instantly, anywhere you can type. Punctuation and cleanup included.
Record mode
Capture a whole meeting or long note. It keeps recording in the background, then transcribes on stop.
Speaker labels
On-device diarization figures out who said what and lays the transcript out by speaker.
Translate, offline
Turn any transcript into 59 languages on demand. The translation models run on the device too.
Flow bubble & keyboard
A floating mic over any app, plus a full voice keyboard. Dictate into messages, notes, anything.
Pick your engine
From an instant built-in model to large multilingual ones. Choose by purpose, download once, run forever.
Dictate into every app you already use.
Scribe is a system keyboard and a floating mic, so it types anywhere you can. No per-app integrations to wire up, no plugins to install.
// the floating Flow Bubble and voice keyboard reach any text field on the system
The audio never gets a chance to leave.
There's no upload step to opt out of. The model lives on your device, so your voice is processed where it's spoken.
You speak
The mic feeds audio straight into a model already downloaded to your device.
It transcribes on-device
Whisper / NeMo / your phone's own engine turns sound into text locally, even in airplane mode.
You keep the text
Edit, label speakers, translate, copy, or share. Nothing was ever sent to a server.
Built for people who can't send their voice to the cloud.
Cloud dictation answers "where does my audio go?" with a compliance badge. Scribe answers it with architecture: it never leaves.
Protect your sources
Transcribe sensitive interviews without a third party ever hearing a second of audio. There is nothing to subpoena from a vendor, because there is no vendor.
Notes that stay confidential
Dictate about patients without routing their words through someone else's cloud. Privacy by architecture, not by a policy page.
Privilege stays privileged
Privileged conversations never leave the device, so there is no copy sitting on a server waiting to be discovered.
Dictate into your editor
Talk comments, commit messages, and prompts straight into Cursor or VS Code, fully offline, even on a plane.
Capture every lecture
Record and transcribe lectures or field interviews with no signal and no upload, then translate the notes on device.
Works at zero bars
Flights, trains, basements, dead zones. Transcription that never waits for a connection it cannot get.
Same job. Different deal.
The most private dictation app is the one that can't betray you.
Scribe has no backend to breach, no logs to subpoena, and no analytics SDK quietly phoning home. Models run on your hardware; the only network call is the one-time, optional model download. Turn on airplane mode and it still works. That's the proof.
Straight answers.
Is Scribe really free?
Yes. Scribe is free and open source. There is no subscription, no trial that expires, and no paywalled features. There is nothing to sell you because there is no server to run.
Does my voice or audio ever leave my device?
No. Transcription runs entirely on your device using a model you download once. Audio is processed in memory and is never uploaded. Switch on airplane mode and Scribe still works, which is the proof.
Does Scribe work offline?
Completely. After the one-time model download, every feature (live dictation, long-form recording, speaker labels, and translation) runs with no internet connection at all.
What languages does it support?
Transcription covers dozens of languages depending on the engine you choose, and on-device translation works across 59 languages. The translation models also run locally on your device.
Which platforms can I use it on?
Android is available in beta today, alongside a macOS menu-bar app and a Windows tray app. An iOS version is coming soon.
How is Scribe different from Wispr Flow or Superwhisper?
Scribe does the same job, fast dictation in any app, but keeps everything on your device, costs nothing, needs no account, and is open source. Most cloud dictation tools send your audio to their own servers to transcribe it.
Do I need an account or an API key?
No account, ever, and no key required. Scribe works the moment you install it. You can optionally add your own API key for a cloud model, but that is off by default and everything stays offline until you turn it on.
Can it tell who said what?
Yes. On-device speaker diarization labels each speaker and lays the transcript out turn by turn, so recordings of meetings and interviews read clearly.
Is Scribe open source?
Yes. The full source is on GitHub, so anyone can audit exactly what the app does with your audio, which is to keep it on your device.
One voice, every screen.
Speak freely.
Keep it yours.
Free, offline, and open. Install Scribe and your words never have to trust a server again.