A Flashy New AI Tool Could Be a Producer's Dream and a Copyright Nightmare

Imagine being able to hear exactly what's under the hood of any piece of recorded music. You upload a file and a few minutes later, a song like "Born to Run" splits apart to reveal its secrets. Each player's mastery is laid bare: There's Bruce Springsteen's isolated vocal take, every murmur and cry heard clearly; Garry Tallent's propulsive bassline; Clarence Clemons' fired-up saxophone solo; and that memorable sprinkling of glockenspiel, courtesy of Danny Federici. Such is the promise of Spleeter, a free, open-source AI tool that was developed and released by the streaming service Deezer late last year. Using a process called source separation, Spleeter splits the audio file of any given song into four new audio "stems," which isolate particular instruments or groups of instruments: vocals, bass, drums, and so on.