Arktan logo

Polymath is a tool that uses machine learning to convert any music library into a music production sample-library, automatically separating songs into stems, quantizing them to the same tempo, analyzing musical structure and converting audio to MIDI.


  • Music source separation with the Demucs neural network
  • Music structure segmentation/labeling with the sf_segmenter neural network
  • Music pitch tracking and key detection with the Crepe neural network
  • Music to MIDI transcription with the Basic Pitch neural network
  • Music quantization and alignment with pyrubberband
  • Music info retrieval and processing with librosa

Use Cases

  • Creating a searchable sample library for music producers, DJs, and ML audio developers
  • Combining elements from different songs to create unique compositions
  • Creating a large music dataset for training generative models

Suited For

  • Music producers
  • DJs
  • ML audio developers