AudioShake wins Sony Demixing Challenge

This summer, Sony launched the first-ever “Demixing challenge.” In this global contest, participants from around the world — including research institutes, private companies, and hobbyist coders — submitted their AI models to run against the same test set so that their results could be compared by SDR score.

AudioShake scored nearly a half decibel higher than the second-place winner, a researcher from Facebook. In third place was a research team from Korea University. TikTok parent company, Byte Dance, was also in the top ten. Particularly notable was AudioShake’s vocal stem, which was a full decibel higher than any other competing team.

What’s most exciting about this contest is all the momentum building in the field of music information retrieval. The uses of sound separation are broad — from music and entertainment experiences, to hearing aids and other audio experiences — so having more researchers tackling these challenges offers the possibility to address real problems that affect people in their daily lives.