Local voice dictation for macOS

Dictate anywhere.
Nothing leaves your Mac.

Hold a key, speak, and your words appear wherever the cursor is — transcribed entirely on-device.

Right Option hold, speak, release
Apple Silicon · macOS 26 · 510 MB, model included
What it does

Fast, private, and it stays out of your way.

A menu-bar app built on a local Whisper model. No cloud, no account, no telemetry — the only time it touches the network is to download its model once.

Talk anywhere

Hold your hotkey in any app — Notes, a browser, Slack, your terminal — and the text lands at the cursor.

Nothing leaves your Mac

Speech is transcribed on-device with Metal acceleration. Fully offline once the model is installed.

Continue, don't repeat

It reads the field first — if you already typed the start of a sentence, it picks up from there instead of duplicating.

Catch both sides of a call

A separate hotkey transcribes system audio — the other person on a Zoom or Meet call — straight into your notes.

Cleans up as you go

Deterministic tidy-up by default; optional on-device grammar, structure, and tone rewrites when you want them.

Learns your words

Watches how you edit inserted text and quietly fixes recurring mis-hears — names, jargon, product terms.

By the numbers

Tuned to feel instant.

Benchmarked on Apple Silicon with the recommended model, greedy decode and a vocabulary hint.

~0.6s
Transcribe 23s clip
<2s
Speak to inserted
100%
On device
0
Accounts or cloud
Install · about a minute

Getting set up.

You were sent a zip. Requires an Apple Silicon Mac on macOS 26 or newer.

  1. Unzip the file
    Double-click the zip you were sent to expand it.
  2. Right-click install.command → Open
    Not a double-click. macOS asks once to confirm an unidentified developer — click Open. It installs the app and launches it.
  3. Grant Microphone & Accessibility
    The setup guide walks you through both. They let FreeKit hear you and type for you — nothing leaves your Mac.
  4. Hold Right Option and talk
    Anywhere. FreeKit lives in your menu bar for settings, models, and the system-audio hotkey.
Two downloads. The full build (510 MB) includes the speech model and works offline immediately. The lite build (4 MB) downloads the model once on first launch. Either way it runs completely on-device afterward. If install.command won't open, drag it into Terminal and press Return.