Open-source software we build on

PlayByPart is made possible by the projects below. We are grateful to their authors. Licence texts are available from each project.

• Demucs (Meta AI Research) – music source separation models and code – MIT.
• PyTorch – BSD-3.
• Audio Spectrogram Transformer (MIT, fine-tuned on AudioSet) via Hugging Face Transformers – BSD-3 / Apache-2.0.
• Basic Pitch (Spotify) – note transcription – Apache-2.0.
• librosa – audio analysis – ISC.
• Verovio – music notation engraving – LGPL-3.0.
• music21 (MIT) – notation parsing – BSD-3.
• Audiveris – optical music recognition, run as a separate program – AGPL-3.0 (source: github.com/Audiveris/audiveris).
• FluidR3 General MIDI SoundFont (Frank Wen) – MIT.
• TinySoundFont / tinysoundfont – MIT.
• FFmpeg – audio conversion, run as a separate program – LGPL-2.1+/GPL-2.0+ (build-dependent; source available from ffmpeg.org).
• Rubber Band Library (within FFmpeg) – GPL-2.0+.
• resvg – SVG rendering – MIT/Apache-2.0.
• pypdfium2 / PDFium – BSD-3 / Apache-2.0.
• ReportLab – PDF generation – BSD-3.
• FastAPI, Starlette, Uvicorn, Pydantic, NumPy, SciPy, soundfile, pretty_midi – MIT/BSD.
• React, Vite, Tailwind CSS, SoundTouchJS – MIT / LGPL (SoundTouch).

Programs under the GPL or AGPL are used unmodified as separate processes; their source code is available from the links above.