Scribe is a fast, 100% private, open-source voice dictation application designed specifically for macOS on Apple Silicon. It converts spoken words to text anywhere you type using the on-device Aether Engine without sending any audio to external cloud servers.

Does Scribe send audio to cloud servers?

No. All core audio transcription happens 100% locally on your Mac using the Apple Silicon Neural Engine. Your voice never leaves your computer.

How do I install Scribe on macOS?

You can download the official DMG installer directly from scribe.dosimple.app or install via Homebrew using the command: brew install --cask bledny1099/scribe/scribe

Is Scribe free and open source?

Yes. Scribe is completely free and licensed under the open-source MIT license. The source code is publicly accessible on GitHub at https://github.com/bledny1099/Scribe.

How does AI refinement and polishing work in Scribe?

Scribe supports intelligent AI modes including Smart Polish (removes verbal clutter and hesitations while preserving speech fidelity), Summary, Executive rephrasing, and Action Items using local models (Apple Intelligence, Ollama) or your own API keys for Claude, OpenAI, or Gemini.

Does Scribe support recording long lectures and meetings?

Yes. Scribe includes a dedicated Lecture Mode designed for 1-2 hour recordings. It records in the background from the menu bar without any floating on-screen overlays, uses acoustic conditioning for large rooms, and automatically exports structured notes to Apple Notes, Obsidian, and Notion.

Can I import audio and video files for offline transcription?

Yes. Scribe can import and transcribe audio and video files (including .mp3, .m4a, .mp4, .mov, and .wav) directly on your Mac with 100% offline on-device processing.

Scribe — Fast, Private Open-Source Voice Dictation for macOS

Scribe

Fast, private dictation & lecture capture for macOS. 100% open-source.

Download for macOS Star on GitHub v2.10.1

$ brew install --cask bledny1099/scribe/scribe

macOS 14+ (Sonoma, Sequoia) · Apple Silicon (M1/M2/M3/M4/M5)

0:00 / 0:00

Engineered for deep flow

From sub-second voice dictation to smart AI polishing, lecture capture, and direct note sync.

AI Intelligence

Smart AI Modes & Refinement

Transforms spoken thoughts with specialized prompt modes. Smart Polish cleans stutters without changing your words, while Executive and Summary modes extract structured meeting briefs.

Smart Polish Summary Executive Brief Action Items

Raw "um, so basically we need to update the raft election timeout, like, right away..."

AI "We need to update the Raft election timeout immediately."

Real-time

Aether Engine

On-device neural inference directly on Apple Silicon. Instantaneous recognition with zero fan spin or battery drain.

Long-Form Capture

Lecture & Meeting Mode

Designed for 2-hour university lectures and multi-hour team standups. Runs quietly from the menu bar with zero screen overlay, so your display stays 100% clean for slides, Zoom, and code.

01:42:15 Distributed Systems Lecture

Acoustics Balanced

System-Wide

One Global Shortcut

Press once to speak, press again to paste directly into whatever application has focus without clipboard pollution.

⌥ Option + S

Push-to-talk or Toggle

Batch Import

Media File Import

Drop existing lecture recordings or Zoom video files for rapid offline batch transcription on Apple Silicon.

Drop recordings

.mp3 .m4a .mov .mp4

Context-Aware

Dynamic App Context

Detects the frontmost application and primes technical vocabulary specifically for Claude Code, Cursor, Antigravity, and terminals.

Claude Code

/compact, MCP, subagents

Cursor

Composer, @Files, diff

Antigravity

Browser agent, artifacts

Terminal

grep, zsh, worktree

Export

Direct Note Sync

Automatically formats lecture notes and transcripts into clean Markdown with timestamps, piping directly into your notes.

Apple Notes

Obsidian

Notion

Local Neural Engine

Zero Telemetry

GPL-3.0 Open Source

No Audio Leaves Mac

Why we built Scribe

We dictate all day

Mostly for prompting coding agents, drafting documents, and sending quick chat replies. Nothing out there was fast, private, and customizable enough, so we built Scribe and open sourced it.

Explore on GitHub

Copied to clipboard!