Verbalog for macOS

Verbalog

Private meeting notes, transcribed on your Mac

Verbalog records your meetings, transcribes them, works out who said what, and lets you correct the result. Everything happens on your Mac, with local AI models: your audio never leaves it, and your transcripts only leave if you choose to send them somewhere. No account, no subscription, no tracking.

Download for macOS · Source code on GitHub

What it does

  • Records your microphone and the call’s audio (Teams, Zoom, Meet…), a meeting room, or both
  • Transcribes with Whisper, accelerated by Apple’s MLX, in French, English, German, Dutch, Spanish, Italian and Portuguese, alone or mixed in one meeting. You tick the languages spoken: with French alone, a French voice can’t be mistaken for English
  • Separates the voices with pyannote, then lets you name each speaker from short audio extracts
  • Takes timestamped notes while you listen: actions, decisions, questions, bookmarks, private notes
  • Mutes your mic in one keystroke when you need an aside to stay off the record
  • Exports Markdown and JSON, ready for your own notes or for the LLM of your choice
  • Prepares a summary through Claude, ChatGPT, Gemini, Mistral or any other LLM, only when you ask: names are pseudonymized first, and you see exactly what would be sent

Privacy by design

Transcription and speaker separation run entirely on your Mac. The network is only used to install the app’s components and models on first launch, and to send text to an LLM when you choose to. There is no analytics or telemetry, including from the libraries Verbalog relies on.

Audio is deleted as soon as the transcript is validated. Voiceprints, if you use them, are encrypted with a key kept in your macOS Keychain.

Free and open source

Verbalog is free software, released under the GNU General Public License v3. The whole code is on GitHub: you can read it, check what it does, build it yourself, report a bug or contribute.

Releases are signed and notarized by Apple. If Verbalog saves you time, you can support its development with a voluntary contribution of the amount you choose. It doesn’t unlock anything: the app is complete for everyone.

Requirements

  • A Mac with Apple Silicon (M1 or later), running macOS 14 or later
  • An internet connection for the first launch, and about 4 GB of disk space for the app’s components and models
  • Optional: a free Hugging Face account, to download the speaker separation model

Learn more