Murmur

Murmur

Voice to text in milliseconds.

No cloud. No API keys. Hold a hotkey. Speak. Release.

Install

Currently available for macOS (Apple Silicon). Windows coming soon.

Homebrew Recommended 
brew tap anubhavitis/murmur && brew trust --tap anubhavitis/murmur && brew install --cask murmur
Script 
curl -sSL https://raw.githubusercontent.com/anubhavitis/murmur/main/install.sh | sh

On first launch, Murmur downloads the tiny.en model (~74 MB). It auto-starts on login and restarts on crash.

Upgrade

Updating to a new version of Murmur.

Homebrew 
brew trust --tap anubhavitis/murmur && brew update && brew upgrade --cask murmur
Script 
curl -sSL https://raw.githubusercontent.com/anubhavitis/murmur/main/install.sh | sh

Re-grant permissions after upgrade

Since Murmur is not notarized, macOS ties permissions to the binary's code signature. When the binary changes after an upgrade, your existing permissions become invalid.

After upgrading, go to System Settings > Privacy & Security and for each of the three permissions (Input Monitoring, Microphone, Accessibility):

  1. Remove murmur from the list
  2. Relaunch Murmur
  3. Re-grant the permission when prompted

How it works

1

Hold

Press and hold Right Option (or Caps Lock)

2

Speak

Talk naturally — Murmur captures audio in real time

3

Release

Text appears in your clipboard or pastes at cursor

No cloud. No API keys. Everything runs on your machine.

Features

☰

Menubar app

Lives in your system tray. Zero windows, zero distractions.

🔒

Fully offline

Runs locally on your machine. No data ever leaves your device.

⌨

Hold-to-record

Right Option or Caps Lock. Natural push-to-talk.

⬇

Three tiers

Fast and Accurate. Pick your speed-vs-quality tradeoff from the menu.

📋

Clipboard or paste

Copy to clipboard, or paste directly at cursor position.

🌐

98+ languages

Multilingual transcription on the Accurate tier. 25 languages on Fast.

🔈

Audio feedback

Beep on record start and stop. Always know the state.

⚡

Auto-start

Launches on login via Launch Agent. Always ready.

Permissions

Grant these in System Settings > Privacy & Security

⌨

Input Monitoring

Detects hotkey press (Right Option / Caps Lock)

🎤

Microphone

Captures audio for transcription

⚙

Accessibility

Simulates Cmd+V for paste-at-cursor

Look for murmur in each permission list and toggle it on.

Tiers

Pick a tier from the menubar. Models download automatically.

Fast

Default

Lowest latency. Runs Parakeet v3 on the Neural Engine, with Whisper large-v3-turbo as fallback.

Download574 MB
Languages25
EngineFluidAudio / Whisper

Accurate

Best quality. Uses Whisper large-v3-turbo, GPU-accelerated via Metal.

Download1.6 GB
Languages98+
EngineWhisper

Models are stored at ~/.murmur/models/. Switching tiers downloads the model in the background with zero downtime.