WhisperBar 🎤

Local voice-to-text that lives in your macOS menubar. No API key, no cloud, completely free.

Hold Ctrl+Shift+R → speak → release → text is pasted wherever your cursor is.

---

Quick Install

Step 1 — Install ffmpeg

brew install ffmpeg

Don't have Homebrew? Install it first: https://brew.sh

Step 2 — Clone the repo

git clone https://github.com/vbiff/whisper_app.git
cd whisper_app

Step 3 — Install Python dependencies

pip3 install -r requirements.txt

Step 3.5 — Install Local Whisper medium

1. Создать папку кэша

mkdir -p ~/.cache/whisper

curl -L "https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt" -o ~/.cache/whisper/medium.pt

First run will download the Whisper model (~145MB). This takes a minute.

Step 4 — Run

Double-click WhisperBar.command in Finder.

Or from terminal:

python3 whisper_menubar.py

A 🎤 icon will appear in your menubar (top-right corner).

---

Usage

Action	Result
Hold Ctrl+Shift+R	🔴 starts recording
Release Ctrl+Shift+R	⏳ transcribes → pastes text
Click 🎤 → Quit	stops the app

Works in any app — terminal, browser, text editor, Slack, etc.

Language: auto-detected (Russian, English, or any language).

---

Auto-start on Login

Run once to make WhisperBar launch automatically when you log in:

bash install_autostart.sh

That's it. WhisperBar will start silently in the background every time you turn on your Mac.

To disable:

launchctl unload ~/Library/LaunchAgents/com.whisperbar.launch.plist

---

Permissions

On first run, macOS may ask for two permissions:

• Microphone — needed to record your voice
• Accessibility — needed to paste text (Cmd+V simulation)

Go to: System Settings → Privacy & Security and enable both for Terminal (or your Python).

---

Model sizes

Edit MODEL_SIZE in whisper_menubar.py to trade speed for accuracy:

Model	Size	Speed	Accuracy
tiny	~75MB	fast	ok
base	~145MB	fast	good ✅
small	~465MB	medium	better
medium	~1.5GB	slow	great
large	~3GB	slow	best

Default is base — good enough for most use cases.

---

Requirements

• macOS
• Python 3.8+
• ffmpeg (brew install ffmpeg)

---

License

MIT