Dictation that stays on your Mac.
Hold Option, speak, and release to write in the app you’re using. Free and open source, with no account or subscription.
Steno 1.0.1 · macOS 13+ · Apple silicon
Works offline, from the first dictation
Steno includes whisper.cpp, the Small English speech model, and Silero voice detection. Everything you need to dictate is included in the download. Your recordings and text are processed on your Mac.
You can download Medium English or Large V3 Turbo from Settings → Speech model. Downloads need internet access and are checked against the published file before they’re installed; dictation works offline once the model is installed. Start with Small and compare models on your own recordings.
Check the final text. Small can mishear words, and quiet speech after a long silence can lose its opening words.
- Requires
- Apple silicon · macOS 13 or later
- Runs
- whisper.cpp on your Mac
- Bundled
small.en, Silero VAD- Optional
medium.en,large-v3-turbo(downloads need internet)- Network
- Dictation works offline; downloads and web links need internet
- License
- MIT; no account or API key
From speech to text
Microphone
whisper.cpp
local
Local cleanup
word corrections · text shortcuts · cleanup style
Insertion
typing · Accessibility · clipboard paste
The app you were using
Hold Option and speak
Hold Option to record and release to finish. For hands-free recording, press F18 to start and again to stop. Choose a different key in Settings → Recording. Keyboard shortcuts that use Option, such as Option+Arrow, don’t start a recording. Cancel discards the current recording, including while Steno is transcribing.
Release to transcribe
Steno transcribes the recording on your Mac, then applies your word corrections, text shortcuts, and cleanup rules. You can turn on a live preview while you speak; the words may change before the recording is finished.
Continue in your editor
Steno tries typing, Accessibility, and clipboard paste in the order you choose in Settings → Text output. Terminal, iTerm2, Warp, and remote-desktop apps use paste first. Steno copies the text instead when the cursor is in a password field or focus moved while it was transcribing. If Steno reports Copied, press ⌘V.
Find past dictations in History
History keeps and shows your last 1,000 completed dictations, including the original and cleaned text, app, time, and insertion result. Copy an entry, paste it again, rerun cleanup and restore the original, delete it, or delete all history.
Made for Apple silicon Macs
Requirements
- Apple silicon Mac
- macOS 13 or later
- Intel Macs are not supported
Permissions
- Microphone — to record
- Accessibility — to insert text into the app you were using
- Input Monitoring — for shortcuts while another app is focused
Grant these permissions in System Settings → Privacy & Security. You can check their status in Steno under Settings → Permissions.
While Hold Option to talk is on, Steno notices key presses and mouse clicks in any app to tell a shortcut such as Option+Arrow from a dictation. It uses only the fact that a key or button was pressed, never which one.
Demo
A 39-second silent demonstration of Steno dictation on macOS.
Get Steno
Steno 1.0.1
Free, open-source dictation for Apple silicon Macs running macOS 13 or later.
Ready to dictate
The speech engine and Small English model are included. Install the app, grant permissions, and start recording.
Updating
There is no in-app updater; quit and replace the app in Applications.
- Open the latest stable release and download its DMG.
- Drag Steno.app to Applications.
- Launch and grant Microphone, Accessibility, and Input Monitoring.
- Click into a text field, hold Option, speak, release; press ⌘V if Steno reports Copied.
To build from source, clone the repository, build the pinned whisper.cpp runtime, and run Steno from Xcode.
git clone https://github.com/Ankit-Cherian/steno.git
cd stenoSettings, History, Insights
Choose your recording shortcut, speech model, and text output preferences in Settings. Most changes wait until you choose Save changes; Discard restores your saved settings. Light, Dark, and System appearance, along with your accent color, save immediately.
Add preferred spellings and shortcuts for phrases you use often, across apps or in a specific app. Cleanup keeps ambiguous phrases by default; Aggressive removes supported fillers. You can also have Steno pause media while you record on macOS 15 or later. It pauses only an app it can confirm is playing and resumes only what it paused, so media you paused yourself stays paused.
Search past dictations in History. Insights shows your activity over time using dates, app identifiers, and counts, without storing transcript text or audio.
Questions
How do I get Steno 1.0?
Download the DMG from the latest GitHub release and drag Steno to Applications. The speech engine, Small English model, and Silero voice detection model are included.
How do I update Steno?
Quit Steno, download the latest DMG, and replace the app in Applications. There is no in-app automatic updater.
Do I need to download a model first?
The app includes Small English, so you can start dictating after granting permissions. The first recording loads the model; later recordings reuse it when possible. If you build from source, follow Development Setup to install the runtime and models.
Can I use a different model?
Yes. Open Settings → Speech model to download Medium English or Large V3 Turbo. Save or discard pending settings changes before switching models. Downloads need internet access and are checked before they are installed; downloaded models can be removed or downloaded again, and installed models work offline. Compare models on your own recordings to find one that suits your Mac.
What stays local?
Your recordings, transcription, and text cleanup stay on your Mac. History stores completed transcripts locally. Insights stores dates, app identifiers, and counts, without transcript text or audio. Optional model downloads use the internet. This website uses Vercel Analytics; dictation in the app does not.
Which Macs are supported?
Steno runs on Apple silicon Macs with macOS 13 or later. Intel Macs are not supported.
How do recording and live text work?
Hold Option to record and release to finish, or press F18 to start and stop hands-free recording. You can change that key in Settings. Stop finishes a recording; Cancel discards it, including while Steno is transcribing. Keyboard shortcuts that use Option, such as Option+Arrow, do not start a recording. The optional live preview may change before the final text is ready. Steno copies the text instead of inserting it when the cursor is in a password field or focus moved while it was transcribing. If Steno reports Copied, press Cmd+V in your text field or copy the transcript from History.
Does Steno read the text in my editor?
Only if you turn on nearby-text continuation. It reads a limited range around your selection to adjust spacing and capitalization in supported English fields in Apple Mail, Notes, and TextEdit. It skips secure or unsupported fields and targets it cannot confirm. Nearby text is never saved or sent to the speech model.
Is Steno free and open source?
Yes. Steno is MIT licensed, with no subscription, account, or API key required. The source is on GitHub.