VoiceMD - Voice Analysis Application

A modern, offline voice analysis tool that predicts speaker characteristics based on acoustic features.

Features

• 🎨 Modern Interface - Clean, intuitive GUI with modern design
• 🔄 Multi-Model Support - Switch between different trained models in real-time
• 💾 Fully Offline - No internet required after initial setup
• 🎤 Multiple Formats - Supports WAV, MP3, OGG, FLAC, M4A
• 📦 Easy Installation - Simple pip install, models auto-download

Installation

Requirements

• Python 3.8 - 3.13 (Python 3.14+ not yet supported - dependencies numba/llvmlite don't support 3.14. Use Python 3.13 or earlier)
• ~500 MB disk space (including models and dependencies)
• Internet connection for first-time model download only

Quick Install (Recommended - No Git Required!)

pip install voicemd-gui

Then run:

voicemd-gui

Models (~4.4 MB) will automatically download from GitHub Releases on first launch.

Alternative Install (From GitHub)

If you want the latest development version:

pip install git+https://github.com/Honey181/voicemd.git

Manual Install (From Source)

1. Download the repository as ZIP from GitHub
2. Extract it
3. Open terminal in the extracted folder
4. Run: pip install .
5. Run: voicemd-gui

Manual Installation

If you prefer to run from source:

# Clone the repository
git clone https://github.com/Honey181/voicemd.git
cd voicemd

# Install dependencies
pip install -r requirements_app.txt

# Run the app
python app_gui.py

Models will auto-download on first run, or manually run:

python download_models.py

Usage

1. Select a model from the dropdown (Small Dataset or CommonVoice)
2. Browse for an audio file (WAV, MP3, OGG, FLAC, M4A)
3. Click "Analyze Voice"
4. View results instantly

Uninstallation

Remove the Application

pip uninstall voicemd-gui -y

Remove Downloaded Models (Optional)

The model files (~4.4 MB) are stored separately and will remain after uninstalling. To completely remove them:

Windows:

Remove-Item -Recurse -Force "$env:USERPROFILE\.voicemd"

macOS/Linux:

rm -rf ~/.voicemd

Location of models:

• When installed via pip: ~/.voicemd/models/ (or %USERPROFILE%\.voicemd\models\ on Windows)
• When running from source: Project root directory

Multi-Model Support

VoiceMD includes two trained models that can be switched at runtime:

• Small Dataset Model - Faster, good for general use
• CommonVoice Model - More robust, handles diverse accents

Switch between models using the dropdown in the app. No restart required!

Troubleshooting

Models Won't Download

Download manually:

1. Go to https://github.com/Honey181/voicemd/releases
2. Download both .pt model files
3. Place them in the project root directory

Or use the download script:

python download_models.py

Import Errors

Update dependencies:

pip install --upgrade -r requirements_app.txt

Audio Loading Issues

The app uses soundfile/librosa (no FFmpeg required). If you still get errors:

• Windows: choco install ffmpeg
• macOS: brew install ffmpeg
• Linux: sudo apt install ffmpeg

LLVM Version Compatibility (Linux)

If you get an error like llvmlite only officially supports LLVM 20 during installation:

Option 1 - Use Conda (Recommended):

conda create -n voicemd python=3.10
conda activate voicemd
conda install -c conda-forge llvmlite numba
pip install git+https://github.com/Honey181/voicemd.git

Option 2 - Install specific LLVM version:

# Ubuntu/Debian
sudo apt install llvm-14 llvm-14-dev
export LLVM_CONFIG=/usr/bin/llvm-config-14
pip install git+https://github.com/Honey181/voicemd.git

Option 3 - Use older Python:

# Python 3.9 or 3.10 have better llvmlite compatibility
python3.10 -m pip install git+https://github.com/Honey181/voicemd.git

Technical Details

• Framework: PyTorch for model inference
• GUI: Tkinter for cross-platform interface
• Audio Processing: librosa, soundfile, torchaudio
• No FFmpeg Required: Uses soundfile/librosa for audio loading

License

MIT License - Copyright (c) 2020, Jeremy Pinto

See LICENSE file for full details.

Credits

Original Project: VoiceMD by @jerpint (Jeremy Pinto)

Enhanced by: @Honey181 - Modern UI and easy installation

This project builds upon the excellent work of the original VoiceMD team. All credit for the model architecture, training pipeline, and core functionality goes to them.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.