Download MicJot

macOS 14 or later on Apple Silicon (M1 or newer). Intel Macs are not supported. Free; this build has no account service.

Before you start

• A speech model, about 630 MB. MicJot recognises speech on your Mac with Whisper Large v3 Turbo. The model downloads once, in the background, from the first setup screen on; a slim strip in the setup window shows bytes and time left, “ready”, “paused: offline” or a failure with Retry. Dictations you make before it is ready are kept and finished once it is. Until the model is ready, Settings → Dictation shows the same state on its Speech model row, with Cancel during a download and Retry after a failure. Recognition runs on your Mac only; there is no cloud engine to choose.
• Two permissions. Microphone (to hear you) and Accessibility (to type into other apps and read the text near your cursor). Input Monitoring is asked for only if MicJot cannot see the Fn key. MicJot observes keyboard events system-wide and acts only on its own keys; it records nothing. The permissions page has the exact System Settings paths.
• The Fn key. Fn is the default dictation key. If you keep it, set System Settings → Keyboard → “Press 🌐 key to” → “Do Nothing”, so macOS does not open its own emoji picker or dictation. Any other key works too: see the dictation key.

First launch

MicJot's setup window has five stages, Sign in › Permissions › Your key › Try it › Ready, one screen each. The speech model downloads in the background from the first screen on. Under Permissions are the dialogs macOS itself shows along the way; the wording varies a little between macOS versions.

1. Sign in. In a build with no account service, such as MicJot 0.3.1, this screen says “No account needed here” and “This build has no account service. Dictation works without one.”, with Continue. The speech model starts downloading in the background (about 630 MB) the moment you continue; from the next screen on, a slim strip shows the download: bytes and time left, “ready”, “paused: offline” or the failure with Retry.

2. Permissions. Two cards, one line each: Microphone (MicJot listens only while you hold your key or lock hands-free) and Accessibility (it types the text into the app you are using, as if you typed it). Each button opens the macOS prompt, a drawing shows where to click, and the page updates by itself when macOS says yes. Continue needs both. Input Monitoring is not asked for here: MicJot asks for it only if it cannot see the Fn key, on the next screen. MicJot observes keyboard events system-wide and acts only on its own keys; it records nothing. The permissions page has the exact System Settings paths.

   • Microphone. macOS asks “MicJot” would like to access the microphone, with MicJot's explanation of its standby behaviour (the microphone can stay open for 90 seconds after a dictation so your first words are not lost; standby audio is discarded, nothing is stored until you dictate). Click Allow.
   • Accessibility. macOS says “MicJot” would like to control this computer using accessibility features. Click Open System Settings, then switch MicJot on under Privacy & Security → Accessibility. This is what lets MicJot type into other apps and read the text near your cursor; it never reads password fields.

3. Your key. Pick the Globe key (fn, the default), Right Option, Right Command or Control + Option, then press and hold it: on a real press the key lights up and a chip reads “Detected · held 1.2 s” (the hold's own length). If MicJot cannot see the key yet, the screen says so and offers Allow Input Monitoring. If you keep Fn, set System Settings → Keyboard → “Press 🌐 key to” → “Do Nothing”, so macOS does not open its own emoji picker or dictation; the screen names any conflict it finds. You can continue without a detected press and check later in Settings → Dictation, which has the same keys, a live check and a way to record your own key. See the dictation key.

4. Try it. One short practice: hold your key, read the messy line on the screen out loud (“so um can we move the sync to three, no, actually four”), let go, and only what you meant lands in a practice note that is never saved; your speaking speed appears beside it. If the speech model is not ready yet, the screen says MicJot is still getting ready; try again in a moment. “Skip practice” moves on.

5. Ready. Your speaking speed from the practice next to a typing speed (typical figures, labelled as such, if you skipped), then the Free and Pro cards under a Monthly / Annual toggle: the published public-release plans, shown for information, with the line “This build has no account service, so no plan applies on this Mac.” Finish closes setup. “Open MicJot at login” is on by default, so the model is ready before your first dictation of the day.

Finish opens Home: your dictation key as a keycap, a compact row of statistics, and every transcript in one searchable list grouped by day, with Copy on each row and Insert again in its “…” menu. MicJot also lives in the menu bar and the Dock.

Updates

This build has no update feed configured, so it makes no update requests at all. Once a feed exists, MicJot will check one small file at most once a day, with no identifiers, and offer new versions with their release notes; nothing will install without you. Until then, new builds are announced on the changelog.

Uninstall

Quit MicJot, drag it to the Trash, and remove its folders in ~/Library if you want nothing left: the uninstall page lists every path and the commands that reset the permissions you granted.

Licences

MicJot is built on open-source software and speech models; their notices are on the acknowledgements page.