QWhisper

QWhisper is a Qt6-based desktop application that provides real-time speech-to-text functionality using OpenAI's Whisper model. It offers a user-friendly graphical interface for continuous audio transcription with support for multiple output formats and advanced audio processing features.

Guarantee

This software has a 100% Works on My Box Seal of Approval. YMMV.

Features

• Real-time speech-to-text transcription using Whisper models
• Support for multiple Whisper model sizes (tiny, base, small, medium, large, turbo)
• CPU and GPU (CUDA) acceleration support
• Voice Activity Detection with configurable thresholds
• Audio filtering with bandpass filter for improved accuracy
• Multiple audio input sources (microphone, system audio)
• Interactive transcript editing with search functionality
• Multiple output options:
  • Interactive transcript window
  • File output with continuous append
  • Clipboard integration
  • Direct typing to active window (X11/Wayland)
• Automatic model downloading and management
• Persistent configuration settings
• Audio level monitoring and visualization

System Requirements

Required Dependencies

• Qt6 (Core, Widgets, Multimedia, Network)
• CMake 3.16 or higher
• C++17 compatible compiler
• PulseAudio development libraries (Linux)

Optional Dependencies

• CUDA Toolkit (for GPU acceleration)
• X11 development libraries with XTest extension (for window typing on X11)
• For Wayland window typing:
  • (may require xwayland)
  • ydotool (requires ydotoold daemon) or
  • wtype (alternative to ydotool)

Building from Source

Prerequisites

• c++/qt build toolchain

Build Steps

1. Clone or extract the source code
2. Navigate to the qwhisper directory
3. Create and enter build directory:

   mkdir -p build
   cd build

4. Configure the build:

   cmake ..

5. Compile the application:

   make -j$(nproc)

The build process will automatically download and compile the whisper.cpp library. CUDA support will be automatically detected and enabled if the CUDA toolkit is installed.

Installation

System Installation

To install QWhisper system-wide:

sudo ./install.sh

User Installation

To install for the current user only:

./install.sh

The installer will:

• Copy the executable to the appropriate bin directory
• Install the application icon and desktop integration files
• Set up documentation and menu entries

Running the Application

After installation, QWhisper can be launched:

• From the command line: qwhisper
• From your desktop environment's application menu
• By clicking the desktop icon

To run without installation from the build directory:

cd build
./qwhisper

Configuration

QWhisper stores its configuration in ~/.config/qwhisper/config.json following the XDG Base Directory specification. Whisper models are downloaded to ~/.local/share/qwhisper/models/ by default.

First Run

On first launch, QWhisper will prompt to download a Whisper model if none are found. The application supports automatic downloading of all official Whisper models from the Hugging Face repository.

Keyboard Shortcuts

• Ctrl+F: Search within transcript
• Ctrl+S: Save transcript
• Ctrl+O: Open saved transcript
• Ctrl+E: Export transcript

Uninstallation

To remove QWhisper from your system:

./uninstall.sh

The uninstaller will:

• Remove all installed application files
• Optionally preserve user configuration files
• Optionally preserve downloaded models
• Clean up desktop integration

Troubleshooting

CUDA Issues

If you experience crashes with GPU selection, try:

CUDA_VISIBLE_DEVICES="0,1,2" qwhisper

This limits which GPUs are visible to avoid out-of-memory issues.

Audio Issues

If audio capture is not working:

• Check that your audio device is properly configured in system settings
• Verify PulseAudio is running and accessible
• Try different audio input sources in QWhisper settings

License

This project is licensed under the MIT License. See the LICENSE file for details.