Trace

On-device

Private, offline meeting transcription for your Mac

Transcribe meetings on your Mac with nothing leaving the device. An on-device model, no cloud, no account.

Most meeting transcription tools send your audio to a server to turn it into text. That is fine until the meeting is something you would rather not hand to a third party: a one-to-one, a legal call, anything under NDA.

Trace does the whole thing on your Mac. It records your mic and the system audio as two separate tracks, then transcribes them locally with an on-device speech model. Whisper large-v3-turbo is the default, with two faster Parakeet options. There is no account to create and no upload step in the flow. The models detect the spoken language on their own, and the default covers around 99 languages.

Trace only uses the network to fetch its models, once each. The first time you transcribe, it downloads your chosen speech model (about 950 MB for the default) from Hugging Face and caches it. After that, turn your Wi-Fi off and it still works. Every session writes a folder with the two WAV files plus transcript.json, transcript.md and a small meta.json, all on your disk.

Why on-device matters here

  • No upload: Audio and transcripts stay in a folder on your Mac.
  • No account: Nothing to sign up for, nothing tying recordings to a login.
  • Works offline: After the one-time model download, transcription needs no network.
  • Sandboxed: Trace runs in the macOS App Sandbox and its privacy manifest declares no data collection.
  • You own the output: Plain .wav, .json and .md files you can move, delete or back up yourself.

FAQ

Does Trace work without an internet connection?

Yes, once the speech model is downloaded. The first transcription fetches it (about 950 MB for the default model, less for the faster ones), then everything runs locally with no network needed.

What data does Trace collect?

None. The privacy manifest declares no data collection, and the app is sandboxed. Your recordings never leave the Mac.

Where are my recordings stored?

In a per-session folder on your Mac containing mic.wav, system.wav, transcript.json, transcript.md and a small meta.json. You control them like any other files.

Which Macs are supported?

macOS 14.4 or later on Apple silicon. The local speech and summary models need an Apple silicon chip.

Get Trace

Get private, on-device transcription. One-time purchase, no subscription.

Compare

Related guides