So far, especially for music producers and DJs, splitting a song into separate vocals and instruments has been a headache. Practically, there are lots of ways that can be done but the process is very much time consuming and there is only a handful of convenient software. Often, the results are imperfect. Now, a new open-source AI tool makes things easier and consumes very little amount of time. The AI tool is named as Spleeter.
Spleeter was developed by music streaming service Deezer for its personal research purposes. A few days back, Deezer released it as an open-source package by putting the code up on Github for anyone who wants to download and use it.
All you have to do is feed Spleeter with an audio file and it splits it into two, four, or five separate audio tracks. These audio tracks are known as stems. As you can imagine, the results are not perfect but they can be eminently used. Moreover, Spleeter does this real fast.
According to the makers, when it runs on a dedicated GPU, it can split audio files into four stems 100x faster than real-time.
Deezer posted in a blog post: “Spleeter is a neat tool, but in no way do we claim to have “solved” source separation. Hundreds of researchers and engineers working for decades have made the advances and built the tools on which Spleeter is based.”
Led Zeppelin – “Whole Lotta Love”
Led Zeppelin (Without Music)
Led Zeppelin (Without Vocal)
You can check some sample examples of what this little AI tool can do, you can check out this sample generated by Andy Baio from Waxy. He said, “sometimes get a robotic autotuned feel, but the amount of bleed is shockingly low relative to other solutions.”