MDXNet

Ultimate Vocal Remover powered by MDX Net

MDXNet is a high-quality vocal separation tool that uses the MDX Net architecture. It leverages GPU acceleration (when available) and multi-threaded processing to deliver fast and efficient separation of vocals from audio files.

Table of Contents

• Features
• Installation
• Downloading Models
• Usage
  • Python API
• Contributing
• License
• Acknowledgments

Features

• High-Quality Vocal Separation: Utilizes MDX Net for precise separation.
• GPU Acceleration: Automatically uses GPU if available.
• Multi-Threaded Processing: Optimized for faster processing on multi-core systems.

Installation

Install MDXNet directly from GitHub using pip:

pip install git+https://github.com/TheNeodev/mdxnet.git

Make sure you have Python 3.7 or later installed.

Downloading Models

MDXNet requires pre-trained models to operate. Download the required models from the releases page:

• Model Repository:
  Download Models

After downloading, place the model file (e.g., uvr_models.onnx) in a directory of your choice, and update the model path accordingly in your configuration.

Usage

MDXNet can be used within your Python scripts.

Python API

Below is an example of how to use the Python API for vocal separation:

from mdxnet import MDXProcessor

# Define your model parameters
model_params = {
    # Customize model parameters here
    # e.g., "param1": value, "param2": value,
}

if __name__ == "__main__":
    # Initialize the processor with the model path and parameters.
    processor = MDXProcessor(
        model_path="./uvr_models.onnx",  # Update this path to your downloaded model
        model_params=model_params,
        processor=0  # Set processor index (use 0 for CPU, or specify GPU device index)
    )

    # Process the audio file to separate vocals and instrumental tracks.
    main_path, invert_path = processor.process(
        input_path="./Test.mp3",   # Path to the input audio file
        output_dir="./output",     # Output directory for the separated tracks
        denoise=True,              # Enable denoising (set to False if not needed)
        suffix="Vocals",           # Suffix for the vocal track file
        invert_suffix="Instrumental"  # Suffix for the instrumental track file
    )

    print(f"Separated vocals saved to: {main_path}")
    print(f"Instrumental track saved to: {invert_path}")

Notes:

• Ensure the model file is correctly placed and the path is updated.
• Customize the model_params dictionary based on your specific requirements.

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository.
2. Create a new branch for your feature or bugfix.
3. Make your changes and commit them with clear messages.
4. Submit a pull request detailing your changes.

For major changes, please open an issue first to discuss what you would like to change.

License

This project is licensed under the MIT License.

Acknowledgments

• Thanks to the developers behind MDX Net/UVR for their groundbreaking work.
• Special thanks to all contributors and the community for continuous support.