Polymath

Polymath is an open-source, Python-based tool created by developer Samim that turns any personal music library, or YouTube videos, into a searchable sample library for music production. It runs source separation using the Demucs neural network to split tracks into individual stems such as drums, bass, vocals, guitar, and piano, then uses libraries like librosa and pyrubberband to quantize every stem to a shared tempo and beat grid, so pieces from different songs line up automatically. Beyond stem separation, Polymath analyzes musical structure, key, timbre, loudness, and pitch, and can convert audio into MIDI for use inside a DAW. It also includes similarity search, letting a producer find tracks in their library that share sonic characteristics with a reference clip. The project is aimed at music producers, DJs, and beatmakers who want to build mashups or sample-based tracks from their own record collections, as well as machine-learning practitioners who need a fast way to assemble training datasets of musical stems. Polymath runs locally on Linux, macOS, or Windows (via Docker) and requires Python 3.7-3.10 along with GPU-accelerated dependencies for the separation models, so setup demands some technical comfort. It is a command-line tool rather than a hosted service, with no subscription or web interface.