macOS & Windows · free · open source

Your words stay
on your machine.

Record calls, transcribe files, dictate text. Entirely offline. No keys, no cloud, no subscription.

macOS 12+ & Windows 64-bit · free forever


what it does

Three modes.
One menu bar icon.

Pick the right tool for the moment. Switch anytime from the menu bar.

Scribe

Background call recorder

Runs silently. Captures both sides of a call via dual-channel recording. Mic bleed suppression keeps the channels clean.

  • BlackHole integration for speaker audio
  • Separate in: / out: transcript channels
  • Saves markdown to a folder you choose
Transcribe

File transcription

Drop any audio file. Get a clean timestamped transcript. All processing is local.

  • Supports WAV, MP3, M4A, FLAC
  • Timestamped output
  • Multiple Whisper model sizes to choose from
Dictate

Hotkey speech-to-text

Tap a key, speak, release. Text transcribes and pastes into whatever app is in focus. Works system-wide.

  • Configurable hotkey
  • Auto-pastes to active app
  • Optional VAD skips silence between words

setup

Install once.
Works forever.

01

Download the .dmg

Mount, drag to Applications. No account, no App Store, no subscription. A menu bar icon appears on first launch.

02

Download a Whisper model

Open Settings → Models and click Download next to a model. Models range from 75 MB to 1.5 GB.

03

Pick a mode

Click the menu bar icon. Choose Scribe, Transcribe, or Dictate. Configure once and forget it.

04

Plain markdown output

Transcripts save to a folder of your choice. Grep them. Pipe them. Open in anything. They're yours.


privacy

Private by
architecture.

Not a setting. Not a promise. Audio is processed entirely on your hardware. There is no server.

Read privacy policy →

No cloud

Whisper runs locally. Audio never leaves the device. No API keys required.

No account

Download and run. No signup, no email, no telemetry of any kind.

No subscription

Free permanently. Open source.

One network call

Model download from Hugging Face. One time, user-initiated. No data sent. Verifiable with a packet capture.

security audit table

Quick-reference
for IT & compliance.

QuestionAnswer
Does the app send audio to any server?No — never
Does the app send transcripts to any server?No — never
Does the app require an account or login?No
Does the app collect telemetry or analytics?No
Does the app phone home for updates?No — but "Check for updates" in Settings → Help makes a single opt-in request to api.github.com if the user clicks the button
Does the app require an always-on internet connection?No
Is an internet connection ever required?Yes — once to download a Whisper transcription model, and optionally once to download the Silero VAD model (~2 MB). Both are user-initiated. Optionally when the user manually clicks "Check for updates".
Where is all user data stored?Exclusively on the user's local device
What AI model runs the transcription?OpenAI Whisper (ggml format), running locally
Does the AI model phone home?No — inference is fully offline
What OS permissions are required?Microphone (all platforms), Accessibility (macOS only for paste), Input Monitoring (macOS only for hotkey)
Can the app be used fully offline after setup?Yes

Full policy including OS permissions, data lifecycle, and compliance notes: privacy policy


download

Run it today.

Free and open source. macOS 12+ and Windows 64-bit.

View all releases v0.1.7 · Choose your platform on GitHub