Polymath

Polymath converts entire music libraries into organized, searchable sample libraries by automatically separating songs into stems such as beats and bass, quantizing them to a consistent tempo, and analyzing musical structure and key.

The tool uses machine learning models including Demucs for source separation, sf_segmenter for structure segmentation, Crepe for pitch tracking, and Basic Pitch for audio-to-MIDI transcription. It also aligns and quantizes audio using pyrubberband and extracts detailed audio features with librosa.

Use cases include creating mash-up DJ sets, generating custom music samples, and building large datasets for training generative music models, benefiting music producers, DJs, and machine learning developers.