Hush — Voice-to-Text for macOS

Local voice-to-text using whisper.cpp. Press-and-hold a hotkey, speak, release — text appears at your cursor. No API costs, no data leaves your machine.

Install

Download the latest Hush.dmg from the build folder, open it, and drag Hush to Applications.

On first launch, right-click the app → Open (to bypass Gatekeeper).

Requirements

• macOS 13.0 (Ventura) or later
• Apple Silicon (M1/M2/M3/M4)

Usage

Launch Hush from Applications. A microphone icon appears in your menu bar.

Hold Ctrl+Space, speak, then release — transcribed text is pasted at your cursor.

Menu bar options

• Status — Ready / Recording / Transcribing
• Accessibility / Microphone — permission status with click-to-fix
• Shortcut — shows current hotkey, with option to change
• Language — Auto-detect, English, or Greek

Menu bar icon states

• 🎤 Ready
• ⏺ Recording...
• ⏳ Transcribing...

macOS Permissions

Hush needs two permissions (prompted on first use):

• Accessibility — for global hotkey capture and simulating Cmd+V paste
• Microphone — for audio recording

Go to System Settings → Privacy & Security to manage these. The menu bar shows permission status and lets you click to open Settings directly.

How It Works

1. Global hotkey listener (Quartz CGEvent tap) detects Ctrl+Space
2. AVAudioEngine streams mic audio at 16kHz mono into a buffer
3. On release, audio is transcribed by whisper.cpp (base model, CPU with Accelerate/BLAS)
4. Language auto-detected between English and Greek (or forced via menu)
5. Transcribed text is copied to clipboard and pasted via Cmd+V

Tech Stack

• Swift / SwiftUI — native macOS menu bar app
• whisper.cpp — local speech-to-text (base model, ~141MB)
• AVAudioEngine — microphone recording
• Quartz CGEvent — global hotkey and keyboard simulation
• Accelerate / BLAS — optimized CPU inference on Apple Silicon

Building from Source

# Clone with submodule
git clone --recursive https://github.com/DimitrisTzimikas/hush.git
cd hush

# Build whisper.cpp static libraries
cd whisper.cpp && mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DWHISPER_BUILD_EXAMPLES=OFF -DWHISPER_BUILD_TESTS=OFF -DGGML_METAL=OFF -DBUILD_SHARED_LIBS=OFF
make -j8 whisper
cd ../..

# Generate Xcode project and build
brew install xcodegen
xcodegen generate
xcodebuild -project Hush.xcodeproj -scheme Hush -configuration Release build

Place ggml-base.bin model in the app bundle's Resources or in ~/Library/Application Support/Hush/models/.

Download the model:

mkdir -p ~/Library/Application\ Support/Hush/models
curl -L -o ~/Library/Application\ Support/Hush/models/ggml-base.bin \
  https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.bin