Polymath

What is Polymath

Polymath is an AI-powered tool that converts music libraries into production sample-libraries using machine learning techniques. It separates songs into stems, quantizes them, analyzes musical structure and key, and converts audio to MIDI. This creates a searchable sample library that streamlines the workflow for music producers, DJs, and ML audio developers. It can be used to create new compositions, discover related tracks, and simplify the process of creating a large music dataset for ML training. Polymath supports GPU acceleration and can be run using Docker.

Key Features of Polymath

- Music Source Separation: Utilizes the Demucs neural network to automatically separate songs into stems like beats, bass, and vocals.

- Music Structure Segmentation: Analyzes the musical structure, such as verses and choruses, using the sf_segmenter neural network.

- Pitch Tracking and Key Detection: Employs the Crepe neural network to track pitch and detect musical key information.

- Music to MIDI Transcription: Converts audio to MIDI files using the Basic Pitch neural network, facilitating easier music production.

- Quantization and Alignment: Uses pyrubberband to quantize songs to a uniform tempo and beat grid, enhancing synchronization.

- Musical Information Retrieval: Extracts and processes musical features such as timbre, loudness, and tempo using the librosa library.

- Searchable Sample Library: Creates a fully searchable sample library to streamline workflow for music producers, DJs, and ML audio developers.

- Integration with DAW: Allows effortless integration of separated, analyzed, and quantized music elements into Digital Audio Workstations (DAWs).

- Diverse Use-Cases: Enables users to mix and match elements from different songs to create unique compositions or DJ sets.

- Community Support: Provides a Discord community for users to exchange insights and seek support.

- Compatibility: Supports Python versions >=3.7 and <=3.10, ensuring wide compatibility with various development environments.

- GPU Support: Offers native GPU support through CUDA for faster processing, with detailed setup instructions available.

- Docker Setup: Includes a Dockerfile for easy building and running of the application in a containerized environment.

- Installation Process: Simple installation steps, including cloning the repository and installing dependencies via pip.

- Customizable Features: Settings for neural networks and feature extractors can be adjusted within the Python files to meet specific needs.

- Library Management: Supports automatic downloading and adding of YouTube videos and audio files to the library.

- Quantization Flexibility: Offers options to quantize songs to a specific tempo or their original tempo.

- Advanced Search Capabilities: Allows searching for similar songs based on various musical features, including tempo.

- MIDI Conversion Quality: While current MIDI conversion results may vary, plans are in place to include additional audio-to-MIDI model options.