TypeNo

May 25, 2026 · View on GitHub

A free, open source, privacy-first voice input tool for macOS.

TypeNo hero image

A minimal macOS voice input app. TypeNo captures your voice, transcribes it locally, and pastes the result into whatever app you were using — all in under a second.

Official website: https://typeno.com

Special thanks to marswave ai's coli project for powering local speech recognition.

How It Works

Short-press Control to start recording
Short-press Control again to stop
Text is automatically transcribed and pasted into your active app (also copied to clipboard)
While recording, the overlay shows segmented preview text about once per second; after you stop, TypeNo still runs a final full-file transcription before pasting

That's it. No windows, no settings, no accounts.

Install

Download TypeNo for macOS
Download the latest TypeNo.app.zip
Unzip it
Move TypeNo.app to /Applications
Open TypeNo

TypeNo is signed and notarized by Apple — it should open without any warnings.

Install the speech engine

TypeNo uses coli for local speech recognition.

Prerequisites:

Node.js (LTS recommended — install directly from nodejs.org for best compatibility)
ffmpeg — required for audio conversion: brew install ffmpeg

npm install -g @marswave/coli

This installs the coli CLI itself, but it does not pre-download the speech model.

If Coli is missing, TypeNo will show an in-app setup prompt with the install command.

Node 24+: If you get a sherpa-onnx-node error, build from source:
npm install -g @marswave/coli --build-from-source

First Launch

TypeNo needs two one-time permissions:

Microphone — to capture your voice
Accessibility — to paste text into apps

The app will guide you through granting these on first launch.

On the first actual transcription, coli will also download its speech model into ~/.coli/models/. That is a separate step from npm install -g @marswave/coli.

Troubleshooting: Coli Model Download Fails

The speech model is downloaded from GitHub. If GitHub is inaccessible in your network, the first transcription can fail while coli is downloading the model.

Fix: Enable TUN mode (also called Enhanced Mode) in your proxy tool to ensure all system-level traffic is routed correctly. Then trigger another transcription so coli can retry the model download.

If ~/.coli/models/ contains a partial .tar.bz2 archive from a failed download, delete the leftover files in that directory before retrying.

rm -rf ~/.coli/models

Troubleshooting: Accessibility Permission Not Working

Some users find that enabling TypeNo in System Settings → Privacy & Security → Accessibility has no effect — a known macOS bug. The fix:

Select TypeNo in the list
Click − to remove it
Click + and re-add TypeNo from /Applications

Accessibility permission fix

Option 2 — Build from Source

git clone https://github.com/marswaveai/TypeNo.git
cd TypeNo
scripts/generate_icon.sh
scripts/build_app.sh

The app will be at dist/TypeNo.app. Move it to /Applications/ for persistent permissions.

Usage

Action	Trigger
Start/stop recording	Short-press `Control` (< 300ms, no other keys)
Start/stop recording	Menu bar → Record
Watch incremental transcription	Overlay updates about once per second while processing
Choose microphone	Menu bar → Microphone → Automatic / specific device
Transcribe a file	Drag `.m4a`/`.mp3`/`.wav`/`.aac` to the menu bar icon
Check for updates	Menu bar → Check for Updates...
Quit	Menu bar → Quit (`⌘Q`)

Uninstall

Quit TypeNo from the menu bar
Drag TypeNo.app to Trash from /Applications
Remove the Coli speech engine and cached models:

rm -rf ~/.coli/
npm uninstall -g @marswave/coli