macOS · Windows · Linux

Download WhisperJot

Built natively for each platform. The local speech models download on first use and run entirely on your device — or skip the download with the Jot Cloud engine (opt-in, used only while selected, and audio never stored on our servers).

See pricing · System requirements · Privacy Policy

For this device

macOS

Built for Apple Silicon. Notarized .dmg. Lives in your menu bar.

Create an account Requires macOS 14 Sonoma or later

Windows

Native C#/.NET port. Installer for x64.

Create an account Requires Windows 10 or later

Linux Beta

Debian/Ubuntu package for x64. Tarball and arm64 builds too.

Create an account Also: tar.gz and arm64 builds

After you download

macOS

  1. Open the app — it lives in your menu bar. There's no Dock icon or main window; look for the mic icon near your clock. First launch opens a browser sign-in (use the account you subscribed with), then a setup window walks you through the rest.
  2. Grant Microphone, Accessibility, and Input Monitoring when prompted — these let WhisperJot hear you and type for you.
  3. The speech model downloads once (Jot Local ≈ 2.3 GB or a smaller Jot Local Pro model) — the progress percentage shows next to the menu-bar icon. After that, local dictation runs offline and launches are instant.
  4. Hold fn (or press ⌥ Space) and speak. Pick your engine and recording mode from the menu bar any time.

Windows

  1. Run WhisperJot-setup.exe. In the first days after a new release, SmartScreen may warn about an unrecognized app — choose “More info”, then “Run anyway”.
  2. WhisperJot lives in your system tray (near the clock). First launch signs you in and downloads the speech model once (~2 GB); after that local dictation runs offline.
  3. Press your hotkey and speak. Right-click the tray icon for Settings, including changing the hotkey.

Linux Beta

  1. Install the package: sudo dpkg -i WhisperJot-linux-*.deb on Debian/Ubuntu, or extract the tar.gz anywhere and run it from there.
  2. Run whisperjot from your app launcher or terminal — the speech model downloads once (~2 GB). Needs PipeWire or PulseAudio; on Wayland install wtype for text injection.