Real-Time In-Browser Time Warping for Live Score Following

Caren M, Egozy E. Real-Time In-Browser Time Warping for Live Score Following. J Audio Eng Soc. 2025;73(12):882-889. doi:10.17743/jaes.2022.0244

@article{Caren2025_23116,
  author = {Caren, Matthew and Egozy, Eran},
  title = {{Real-Time In-Browser Time Warping for Live Score Following}},
  journal = {Journal of the Audio Engineering Society},
  volume = {73},
  number = {12},
  pages = {882--889},
  year = {2025},
  month = dec,
  publisher = {Audio Engineering Society},
  doi = {10.17743/jaes.2022.0244},
  url = {https://doi.org/10.17743/jaes.2022.0244}
}

TY  - JOUR
TI  - Real-Time In-Browser Time Warping for Live Score Following
AU  - Caren, Matthew
AU  - Egozy, Eran
T2  - Journal of the Audio Engineering Society
J2  - J. Audio Eng. Soc.
VL  - 73
IS  - 12
SP  - 882
EP  - 889
PY  - 2025
DA  - 2025/12/16
DO  - 10.17743/jaes.2022.0244
UR  - https://doi.org/10.17743/jaes.2022.0244
PB  - Audio Engineering Society
LA  - en
AB  - State-of-the-art score-following systems have largely been implemented exclusively in standalone computer applications. However, web browser–based audio applications offer several advantages over standalone applications, including cross-platform compatibility, straightforward setup, and greater accessibility. The authors present an implementation of a real-time score-following algorithm using online time warping that runs exclusively in the browser, developed as part of ConcertCue, an application that synchronizes content delivery to live musical performances. This online time warping implementation was developed in Python yet runs in the browser without modification, using a framework the authors developed for shuttling audio and control data between native JavaScript and Pyodide, a browser-based Python interpreter. The authors discuss the performance of this implementation and the generalizability of this approach to developing other real-time audio analysis applications using Python in the browser. In addition to releasing the source code, the authors also release the hand-annotated audio alignment dataset used to test the system.
ER  -