Artists have been making ‘stems’ from their tracks available since the early 2000s for fans to remix, often as part of contests. The recent deal between Tracklib and the estate of Isaac Hayes shows that trend continues. Now Deezer has released an open-source tool that could have an interesting impact.
Spleeter focuses on ‘source separation’, using machine-learning technology to separate tracks into their composite stems – for example, bass, drums, vocals, guitar and piano. This isn’t part of Deezer’s streaming service, but rather a tool developed within the company, which it’s now making available for other MIR (Music Information Retrieval) researchers.
“Why release Spleeter?,” it asks in a blog. “Short answer: we use it for our research and think others might want [to] too.” Deezer also made a point of adding a warning in for potential users of Spleeter. “It’s worth pointing out that music mixing is a fine art and that mastering sound engineers are artists in their own rights. Obviously we do not intend to harm their work in any manner or affect anyone’s credit. When you use Spleeter, please do so responsibly.”
Blogger and technologist Andy Baio has been playing with the tool, and posted examples of separation on tracks from Lizzo, Led Zeppelin, Lil Nas X, Marvin Gaye, Billie Eilish and Van Halen. It’s all very impressive on a tech level, with one use being to help people create smoother, better mash-ups and remixes. Yet actually releasing those tracks through pre-licensed services like Deezer (as opposed to user-upload services like SoundCloud and YouTube) can still be a struggle, despite the efforts of startups like Dubset to establish licensing paths for this content.