Licences and credits
Audio to Sheet Music is built on published research and open-source software. Thank you to their authors.
Piano transcription model
Qiuqiang Kong, Bochen Li, Xuchen Song, Yuan Wan and Yuxuan Wang, “High-resolution Piano Transcription with Pedals by Regressing Onset and Offset Times”, IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2021.
- Model checkpoint: Zenodo 4034264, licensed under CC BY 4.0. Used unmodified.
- Code: piano_transcription_inference, MIT License.
Score rhythm, key and hands (PM2S)
Lele Liu, Qiuqiang Kong, Veronica Morfi and Emmanouil Benetos, “Performance MIDI-to-Score Conversion by Neural Beat Tracking”, Proceedings of ISMIR 2022.
- Model checkpoints: Zenodo 10520196, licensed under CC BY 4.0. Used unmodified.
- Code: PM2S, MIT License.
Piano check (PANNs)
Qiuqiang Kong, Yin Cao, Turab Iqbal, Yuxuan Wang, Wenwu Wang and Mark D. Plumbley, “PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition”, IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2020. Used to check that a recording contains piano before it is transcribed.
- Model checkpoint: Cnn14_16k, Zenodo 3987831, licensed under CC BY 4.0. Used unmodified.
- Code: audioset_tagging_cnn, MIT License.
Bulpara's own code turns these models' output into notation (quantising rhythm, finding the metre, splitting the hands and engraving). Neither the authors nor their institutions endorse Audio to Sheet Music.
Software
- music21: engraving and MusicXML. © Michael Scott Asato Cuthbert and cuthbertLab. BSD 3-Clause License.
- OpenSheetMusicDisplay: score display in the App. © OpenSheetMusicDisplay contributors. BSD 3-Clause License.
Playback piano
Upright Piano KW, a Kawai upright recorded by Gonzalo and Roberto Gordo Sáez for the FreePats project, dedicated to the public domain (CC0 1.0).
Example on the home page
Bach, Prelude in C major BWV 846, performed by Kimiko Ishizaka, from the Open Well-Tempered Clavier, dedicated to the public domain (CC0). We used a phone-quality copy of the first 20 seconds.