Speak. It types.

Fully local dictation for Apple Silicon. Hold ⌥⌘, speak, release. No cloud, no telemetry, no Dock icon — just your voice and a tray icon.

View on GitHub Install from source

Features

Everything you need. Nothing you don't.

A menu-bar dictation tool built for speed and privacy — runs entirely on-device using Apple's Neural Engine.

On-Device Privacy

All transcription happens locally via mlx-whisper. Your audio never leaves your Mac — no API calls, no telemetry, no subscriptions. Audit the source →

Hold-to-Talk Hotkey

Default ⌥⌘ in hold-to-talk mode. Switch to tap-to-toggle for any standard accelerator. The recording starts instantly — no UI to click.

Model Selector

Choose between whisper-turbo, whisper-large-v3-turbo, and more from the tray menu. Settings persist across restarts.

No Distractions

LSUIElement = true — no Dock icon, no window, no splash screen. Just a four-state tray icon and a hotkey.

How It Works

Four states. One hotkey.

The tray icon tracks every stage of transcription so you always know what's happening.

Idle

Recording…

Processing…

Done / Error

Install

Up and running in minutes.

Clone, build, and start. The tray icon appears immediately — then hold ⌥⌘ and speak.

git clone https://github.com/griffinwork40/whisper-app.git
cd whisper-app
npm install
npm run build
npm start

Requirements

What you need first.

Apple Silicon (M-series) — mlx-whisper runs on the Neural Engine and GPU. Intel Macs are not supported.
macOS Sequoia 26.x or later — tested on the current release.
Python 3 with mlx-whisper — pip3 install mlx-whisper. Must be importable from the default python3 on PATH.
ffmpeg at /opt/homebrew/bin/ffmpeg — the default Homebrew path on Apple Silicon. brew install ffmpeg.
One pre-cached HuggingFace model — e.g. mlx-community/whisper-turbo (~810 MB). See the README for the one-liner.

Roadmap

What's next.

PRs welcome. See CONTRIBUTING.md to get started.

Code signing + notarization, ship a .dmg
Streaming / partial transcription
Per-app output-mode overrides
Custom dictionary / replacement rules
Settings UI (currently config file only beyond model + output mode)