Live Transcribe GitHub

Private dictation for your Mac

Hold fn, speak and let go. Live Transcribe types what you said at your cursor, in almost any app, with the “um”s removed and the punctuation fixed. Speech recognition and cleanup run on your Mac, so what you say never leaves it.

Download for Mac

Free and open source · Apple silicon · macOS 14 or later

What you say, and what it types

A reply dictated into Slack: hold the shortcut, speak, let go.

You sayIt types
Um, I think we should, uh, push the launch by a weekI think we should push the launch by a week.
Let’s meet on Monday, no wait, TuesdayLet’s meet on Tuesday.
The budget is fifty thousand, I mean sixty thousandThe budget is 60,000.
So, uh, what time does the, um, the train leaveSo, what time does the train leave?

Outputs from the dictation test set (synthetic speech) at the default cleanup level.

Dictation History keeps what you said next to what it typed, so you can copy your original wording. It stays on your Mac, and you can turn it off.

The Dictation History window. A dictation made in Slack is selected: “Um, I think we should, uh, push the launch by a week” was inserted as “I think we should push the launch by a week.” at the Medium cleanup level.

What it does

Settings

Shortcut and cleanup

Hold fn while you speak, or double-tap it to keep listening until you press it again. You can choose another key. Cleanup has four levels: None only applies your snippets and vocabulary, Light fixes punctuation, capitals and misheard words, Medium also removes fillers and follows your corrections, and High also rewords lightly for clarity.

Settings, General tab. Dictation is on, the dictation shortcut is fn, double-tap for hands-free is on and the undo shortcut is Control-Option-Z. Cleanup is set to Medium, out of None, Light, Medium and High.

Snippets

Save text you use often under a short phrase, such as a link, an address or a template. Say the phrase while dictating and the text goes in exactly as saved.

Settings, Snippets tab, listing four snippets: “my calendar link” inserts a calendar URL, “office address” an address, “sign off” a closing, and “standup template” a Yesterday, Today and Blockers template.

Vocabulary

Add names and jargon, with the ways speech recognition tends to mishear them, and they’re spelled your way.

Settings, Vocabulary tab, listing GitHub, Kubernetes (heard as “cooper netties” or “kuber netes”), MLX (heard as “m l x” or “em el ex”) and Nerdstorm (heard as “nerd storm”).

Apps

For each app, choose how dictated text gets in, typed through Accessibility or pasted, and whether it may break across lines. Every app takes line breaks out of the box, so lists, letters and “new line” work in chat apps too. Terminals stay on one line, so a line break never runs a command, and so does any field that only takes one line, such as a search box.

Settings, Apps tab. Under Your settings, Mail is set to Paste with line breaks and Messages to Accessibility on a single line. Under Built in, Terminal pastes on a single line, and a note says 17 more apps have a built-in setting that applies once they’re installed.

Stays on your Mac

Speech recognition and cleanup run locally with Apple’s MLX. There is no account, no cloud service and no telemetry. The app goes online only to download its models from Hugging Face on first launch and, if you let it, to check GitHub for a new version about once a day. Dictation works offline. Your dictation history is kept on your Mac, and you can turn it off.

Install

  1. Download the latest release from GitHub and move Live Transcribe to your Applications folder.
  2. Open it. Setup walks you through microphone access, Accessibility and the fn key setting, while the speech models (about 2 GB) download.
  3. When the menu bar icon shows it’s ready, hold fn in any app and speak.

The app is signed and notarized by Apple, so macOS only asks you to confirm opening an app downloaded from the internet. It keeps itself up to date. You can also build it from source.

Requirements