Earlier in November, Deezer released an open-source project called Spleeter, which had been created by its research team. It’s a tool for ‘source separation’, using machine-learning technology to separate tracks into their composite stems (bass, drums, vocals etc).
It was interesting, but it wasn’t for beginners: you’d need to be a developer or music information retrieval (MIR) researcher to get it up and running. But the point was that developers could take Spleeter and make something more user-friendly, and now someone has done just that.
It’s called Moises, and is the work of developer Geraldo Ramos, in a weekend. “I had the idea of creating a simple service that removes all the friction and processes Deezer’s algorithm on remote servers that can scale according to the traffic,” he explained in a post on the Product Hunt website.
And it really is user-friendly: you can upload a music track or (and this may cause a bit of a stir within the industry) paste in a YouTube link. Moises will then do its thing, separating out the sources, before making them available to stream or download. The quality seems pretty good too, although obviously varying by the source material.
A boon for mash-up makers, although the tracks they make will be subject to the usual copyright environment if and when they are released.