Overhear

Talk faster than you type.

Overhear turns speech into text in every app on your Mac. Always on, listens, transcribes and pastes wherever your cursor is.

Download for macOSView on GitHub

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

Nothing to press.

Most dictation tools ask you to hold a key while you talk. Overhear was built for working from home. It just keeps listening, and the moment you start speaking, the transcription lands wherever your cursor is.

Dictation tools
  1. Focus the field
  2. Hold the key
  3. Speak
  4. Release
Overhear
  1. Speak
Cancel word

Say the word. Take it back.

Misspeak mid-sentence, say “Alexa”*, and the whole sentence is discarded before it reaches your application, even if transcription is already running.

Let’s push the review to Thursday morning… Alexa

Four cancel words ship with the app, and custom ones can be installed if you would rather it listened for something else.

* The default is “Alexa”. After years of disappointment, it felt right to give the name a job it could actually do.

Three things it’s good at.

Long-form transcription

Speech is transcribed in batches as you pause, so you can think out loud for a whole paragraph instead of dictating one sentence at a time and starting over.

Works across your Mac

Text arrives through the clipboard and a simulated ⌘V, so it lands anywhere that accepts a paste: Slack, your editor, a commit message, a browser form.

Multilingual input

Pick the languages you actually speak from the 56 available; Overhear detects which one you’re using. Answer an email in Polish, go straight back to English.

Your voice never leaves your Mac.

Transcription runs locally with faster-whisper, and voice commands with openwakeword. The models download once during setup; everything after that happens offline.

No accountNothing to sign up for.
No serverAudio is never uploaded anywhere.
No telemetryNothing is measured or reported.

Before you install.

Overhear is an early open-source release. Here is everything it needs and everything that happens the first time you run it.

System
macOS 14 Sonoma or later, on Apple Silicon.
Python
Python 3.10 or later must be installed.
brew install python3
Permissions
Microphone and Accessibility, both requested on first launch.
First run
Overhear sets up a local Python environment and downloads the speech models. That’s about 550 MB and a few minutes, once. After that it works offline.
Not code-signed
Overhear isn’t signed yet. After dragging it to Applications, run:
xattr -cr /Applications/Overhear.app
Download for macOSView on GitHub