This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.
Business News Digital
Deezer releases AI tool to split tracks into individual parts
By Andy Malt | Published on Thursday 7 November 2019
Deezer has released an AI tool that can isolate vocals and instruments in music recordings as an open source application for anyone to use. Called Spleeter, the software is able to split tracks into two, four or five ‘stems’.
The streaming service is talking Spleeter up as a research tool, rather than an attempt to re-ignite the mash-up scene, and stresses that it should only be used on tracks where a researcher or remixer has the permission of any relevant copyright owners.
In a blog post discussing the tool, Deezer’s Head Of R&D Manuel Moussallam said the company was releasing the technology to benefit those involved in what is often called ‘music information retrieval’, the process of retrieving information from music for various different purposes. Moussallam hopes the MIR community will get some benefit from having access to “a state-of-the-art source separation algorithm”.
At full capacity he says that Spleeter is able to process tracks at 100 times real-time, making it “a good option to process large datasets”. It is able to split tracks into two parts (vocals and other sounds), four parts (vocals, bass, drums and others sounds), or five parts (vocals, bass, drums, piano and other sounds).
As for why the company has decided to make the software publicly available, Moussallam explains: “One of the hard limitations faced by MIR researchers is the lack of publicly available datasets due to copyright issues. Here at Deezer, we have access to a fairly large catalogue that we’ve been leveraging to build Spleeter. Since we cannot share this data, turning it into an accessible tool is a way for us to make our research reproducible by everyone”.
“On a more ethical standpoint”, he adds, “we feel there should not be an unfair competition between researchers based on their access to copyrighted material or lack thereof. [And] last but not least, training this kind of model requires a lot of time and energy. By doing it once and sharing the result, we hope to save others some trouble and resources”.
If you fancy having a crack at it yourself, you can start ripping apart tracks by downloading Spleeter here. While respecting everyone’s copyrights, obviously.
