Tribute bands are great because even if you might not be able to see your favorite artists live, they recreate that magic to some extent. But what if that tribute band is an AI?
Over The Bridge, a Canada-based non-profit that helps musicians with mental health, has started a new AI-powered project called Lost Tapes of the 27 Club. The organization is using machine learning to break down songs of artists who died at 27, including Jimi Hendrix, Jim Morrison, Amy Winehouse, and Kurt Kobain — and compose original tunes based on their musical stylings.
Check out this well-composed Niravana tribute track called ‘Drowned in the Sun.’
For the project, the Over The Bridge teambroke down songs by these musicians into MIDI samples, so they can train the AI on different parts of tracks like hooks, rhythms, melodies, and lyrics.
Sean O’Connor, the board of director of the organization, said in an interview the Rolling Stones magazine that if you feed whole songs to AI, it gets confused and the final output won’t be great:
So we took 20 to 30 songs from each of our artists as MIDI files and broke them down to just the hook, solo, vocal melody or rhythm guitar and put those through one at a time. If you put whole songs through, [the program] starts to get really confused on what [it’s] supposed to sound like. But if you just have a bunch of riffs, it’ll put out about five minutes of new AI-written riffs, 90 percent of which is really bad and unlistenable. So you start listening through and just finding little moments that are interesting.
He added that, for lyrics, they used a similar process and started by entering few words of the lyrics to the AI and let it process them. Later, they threw in more words so that the algorithm could digest it.
Recreating songs of famous musicians through AI is not new. However, while some songs may sound like artists they’re trying to imitate, they end up being drab tracks on their own. However, process of humans working with AI to create songs will result in wonderful
You can learn more about the project and listen to the whole album here.
Published April 6, 2021 — 09:48 UTC