VibeFlow

Native macOS voice dictation powered by whisper.cpp. Hold Cmd+Ctrl to record, release to transcribe and paste at cursor. Built with C++/Qt6 and Metal GPU acceleration on Apple Silicon.

Inspired by Wispr Flow.

Features

• Hold-to-dictate: Hold Cmd+Ctrl to record, release to transcribe and paste
• GPU-accelerated: whisper.cpp large-v3 model running on Metal (Apple Silicon)
• Multilingual: Automatic language detection — English, Urdu, and 98 other languages
• Frosted glass UI: Floating waveform bubble with liquid glass effect
• System-wide: Works in any app — dictate into TextEdit, VS Code, Safari, Notes, etc.
• Escape to cancel: Press Escape while recording to abort
• Menu bar app: Lives in the system tray, no Dock icon

Demo

Hold Cmd+Ctrl → frosted glass bubble appears at bottom of screen with animated waveform → speak → release → text appears at cursor.

Requirements

• macOS 14+ (Sonoma or later)
• Apple Silicon (M1/M2/M3/M4)
• ~3GB disk space for the whisper large-v3 model
• Homebrew

Quick Start

# Install dependencies
brew install qt@6 cmake

# Clone with submodules
git clone --recurse-submodules https://github.com/skcadri/vibeflow.git
cd vibeflow

# Download the whisper large-v3 model (~3GB)
bash scripts/download-model.sh

# Build, bundle, and sign
bash scripts/build.sh

# Install
rm -rf /Applications/VibeFlow.app && cp -R build/VibeFlow.app /Applications/

Launch from Applications or:

open /Applications/VibeFlow.app

First Launch

macOS will prompt for two permissions:

1. Microphone — click "Allow" (required for recording)
2. Accessibility — System Settings → Privacy & Security → Accessibility → enable VibeFlow

Architecture

┌────────────────────────────────────────────────────┐
│                     App.cpp                         │
│              (state machine controller)             │
│         Idle ←→ Recording ←→ Processing             │
├──────────┬──────────┬──────────┬───────────────────┤
│ Hotkey   │ Audio    │ Whisper  │ UI                 │
│ Monitor  │ Capture  │ Transcr. │                    │
│          │          │          │ ┌───────────────┐  │
│ CGEvent  │ QAudio   │ whisper  │ │ GlassBubble   │  │
│ Source   │ Source   │ .cpp     │ │ ┌───────────┐ │  │
│ Flags    │ (pull    │ (Metal   │ │ │ Waveform  │ │  │
│ State    │  mode)   │  GPU)    │ │ │ Widget    │ │  │
│ polling  │          │          │ │ └───────────┘ │  │
│ @60Hz    │          │          │ └───────────────┘  │
│          │          │          │ TrayIcon            │
└──────────┴──────────┴──────────┴───────────────────┘
         macOS APIs              Qt6 Widgets
    (CoreGraphics, AppKit)

Project Structure

vibeflow/
├── CMakeLists.txt              # Build configuration
├── src/
│   ├── main.cpp                # Entry point
│   ├── App.h / App.cpp         # State machine controller
│   ├── Transcriber.h / .cpp    # whisper.cpp wrapper
│   ├── AudioCapture.h / .cpp   # Mic recording (pull-mode QAudioSource)
│   ├── HotkeyMonitor.h / .mm   # Cmd+Ctrl detection (CGEventSourceFlagsState polling)
│   ├── TextPaster.h / .mm      # Clipboard + Cmd+V paste simulation
│   └── ui/
│       ├── GlassBubble.h / .mm # Frosted glass floating pill
│       ├── WaveformWidget.h/.cpp# Animated 24-bar equalizer
│       └── TrayIcon.h / .cpp   # Menu bar icon
├── deps/
│   ├── whisper.cpp/            # Git submodule
│   └── qt-liquid-glass/        # Git submodule
├── resources/
│   └── Info.plist              # macOS bundle metadata
├── scripts/
│   ├── build.sh                # Build + bundle + sign pipeline
│   └── download-model.sh       # Model download helper
├── models/                     # Model files (gitignored)
│   └── ggml-large-v3.bin
└── AGENTS.md                   # Detailed codebase guide for contributors

Current Status

Working:

• Hotkey detection (Cmd+Ctrl hold/release via polling)
• Whisper model loading with Metal GPU acceleration (3GB large-v3)
• Frosted glass bubble UI with waveform animation
• State machine (Idle → Recording → Processing → Idle)
• Menu bar tray icon
• Text paste at cursor simulation
• macOS app bundle with bundled Qt frameworks

Not working:

• Audio capture delivers silence — QAudioSource reports active with no errors, but macOS TCC microphone permission doesn't persist across ad-hoc signed rebuilds. See AGENTS.md for detailed diagnosis and suggested fixes.

Building from Source

Manual Build

cmake -B build \
    -DCMAKE_PREFIX_PATH=$(brew --prefix qt@6) \
    -DCMAKE_BUILD_TYPE=Release

cmake --build build -j$(sysctl -n hw.ncpu)

The resulting app bundle is at build/VibeFlow.app.

Build Script

scripts/build.sh handles the full pipeline:

1. CMake configure + build
2. macdeployqt to bundle Qt frameworks
3. install_name_tool to fix Homebrew rpath references
4. codesign to re-sign everything (ad-hoc)
5. Copy whisper model into app bundle Resources

License

MIT