Skip to content

Hermes logo

Hermes

Local-first, privacy-focused meeting transcription for macOS.
No cloud services. No data leaves your device.

Latest Release Platform Apple Silicon

Hermes captures audio from your calls — Zoom, Google Meet, Microsoft Teams, FaceTime, or anything that outputs system audio — and your microphone simultaneously. It transcribes both streams locally using WhisperKit and displays a live transcript in a floating overlay that hovers above your call window.

Think of it as a personal, offline Granola — transcription only, no data leaves your Mac, ever.

  • Dual-Stream Capture

    System audio (remote participants) via Core Audio Taps and microphone (you) via AVAudioEngine. Free speaker diarization — "Me" vs "Them" — with no ML speaker identification needed.

  • On-Device Transcription

    WhisperKit running on Apple Neural Engine with the large-v3 model. Nothing leaves your Mac. No API keys, no subscriptions.

  • Floating Overlay

    A transparent panel that hovers above your call window. Collapses to a tiny pill icon when you don't need it. Always on top, never in the way.

  • Completely Private

    No network calls, no analytics, no telemetry, no cloud anything. Audio is captured, transcribed, and stored on your Mac — period.

  • Global Hotkey

    Start and stop recording from anywhere with Cmd+Shift+R. No need to switch windows.

  • Session History

    All transcripts are persisted locally via SwiftData. Browse past sessions and export as Markdown.

  • Installation

    Download the DMG or build from source.

  • Usage

    Learn how to record, transcribe, and manage sessions.

  • Architecture

    How Hermes captures audio and transcribes it.

  • Privacy

    What data Hermes collects (nothing) and where it goes (nowhere).