A private macOS dictation utility
Local Dictation
Local Dictation is free. Speak naturally and put the words into the app you are already using—speaking can be roughly 5× faster than typing, depending on your pace and workflow. Everything runs on your Mac: no account, no API key, no network connection, and nothing sent anywhere.
- System
- macOS 14 or later
- Macs
- Apple Silicon and Intel
- Language
- English only in the first release
- Download
- About 140 MB, because the speech model is included
- Testing
- Tested only on Apple Silicon, specifically an M3 Pro Mac. Intel Macs are untested.
No Terminal or model setup needed
Install Local Dictation
- Download the DMG. Use the download button below this list.
- Drag Local Dictation to Applications.
- Approve the app once. This release is signed but not notarized by Apple, so macOS will say it cannot verify the developer. That warning is expected. Control-click or right-click Local Dictation in Applications, choose Open, then confirm Open. If macOS blocks it anyway, open System Settings → Privacy & Security, scroll down, and choose Open Anyway. You only do this once.
- Grant the permissions the app requests. macOS will guide you through Microphone and Accessibility access needed for dictation. The app may also request Input Monitoring, a macOS permission used for the global shortcut.
- Begin dictating. Use Local Dictation in the app where you want your words to appear.
Recently updated
What's new
0.2.1 — Sep 12, 2026
- Press Escape to cancel a recording.
- Faster transcription: text appears sooner, especially when your Mac is busy.
Older versions
0.2.0 — Sep 10, 2026
- Settings window with custom shortcuts.
- Recent Transcripts menu keeps your last 30 results.
- More accurate speech model (base.en).
- Silent recordings no longer paste sound labels like [BLANK_AUDIO].
0.1.0 — Sep 9, 2026
- First public release.
Private by design
Nothing leaves your Mac
Audio is transcribed on your own machine and the recording is deleted straight afterwards. Your last 30 transcripts are saved locally so you can find one again, in a file only your account can read. Clearing them is one click, and nothing is ever uploaded.
Ready after install
Speech recognition is bundled
The app includes the open Whisper runtime and the ggml-base.en.bin model, so there is no paid cloud service behind it. You do not need to download a model, sign in, or configure anything.
Start and stop
Your shortcut, or the menu
Press right Option or the middle mouse button to start, and again to stop. You can record your own shortcut in Settings instead. The menu bar does the same job on its own, reading Stop Dictation while it is listening.
If you lose a paste
Recent Transcripts
Every result is kept before it is pasted, so a paste that lands in the wrong window is never lost. Open Recent Transcripts in the menu and select one to copy it back. A recording that picked up only silence or background noise is discarded rather than pasted.