open.unmix.app - towards audio separation on the edge
Audio separation has recently become a mature technology, with state of the art reaching performance that permits wide audience application. While recent research continues to develop sophisticated methods to make results even better, we present our recent work in bringing the reference open-unmix model to the edge. This results in open.unmix.app, a web-based platform for audio separation. We demonstrate two contributions: First, we report on the difficulties of bringing a state-of-the-art deep learning-based model to javascript using TensorFlow.js including all necessary pre-and postprocessing. Solving them enables performing separation and enhancement of music and speech signals directly in the browser. Second, we present a platform that exploits this web-separation technology and allows users to share separated multitrack audio on the web using a custom, embeddable multi-track player.
@inproceedings{2021_41,
abstract = {Audio separation has recently become a mature technology, with state of the art reaching performance that permits wide audience application. While recent research continues to develop sophisticated methods to make results even better, we present our recent work in bringing the reference open-unmix model to the edge. This results in open.unmix.app, a web-based platform for audio separation. We demonstrate two contributions: First, we report on the difficulties of bringing a state-of-the-art deep learning-based model to javascript using TensorFlow.js including all necessary pre-and postprocessing. Solving them enables performing separation and enhancement of music and speech signals directly in the browser. Second, we present a platform that exploits this web-separation technology and allows users to share separated multitrack audio on the web using a custom, embeddable multi-track player.},
address = {Barcelona, Spain},
author = {Stöter, Fabian-Robert and Jacintho, Maria Clara Machry and Vaz, Delton de Andrade and Uhlich, Stefan and Mitsufuji, Yuki and Liutkus, Antoine},
booktitle = {Proceedings of the International Web Audio Conference},
editor = {Joglar-Ongay, Luis and Serra, Xavier and Font, Frederic and Tovstogan, Philip and Stolfi, Ariane and A. Correya, Albin and Ramires, Antonio and Bogdanov, Dmitry and Faraldo, Angel and Favory, Xavier},
month = {July},
pages = {},
publisher = {UPF},
series = {WAC '21},
title = {open.unmix.app - towards audio separation on the edge},
year = {2021},
ISSN = {2663-5844}
}