Music with Artificial Intelligence

Pretty useful.

Pretty great.

Amid all the worry about artificial intelligence killing jobs, replacing the human touch, and taking over the world, today's post is a first-person report on how AI helps create music.  This guest post by Rick Millward does not describe AI as a Frankenstein monster, destroying creativity. Quite the opposite. The post describes the experience of finding AI to be a useful tool for music creators.

Rick Millward is a songwriter, musician, music producer, and performer. He worked in Silicon Valley, then Nashville, and now has retired in Southern Oregon, where he performs primarily in wine venues. His most recent album is Loveland. It and his other music are available on Spotify and other streaming services.

Guest Post by Rick Millward

I thought it would be interesting to share a personal experience with an AI application specifically targeted to music production. 

When a song is recorded, the final product is a stereo, or two-channel “mix.” All the various instruments are encoded together. Once this is done, it can’t be modified, or as we say in the biz, it’s a “master.” The source recordings are either separate digital files, or prior to computers, on magnetic tape, where each instrument has its own separate track. These tapes were big brothers to cassette tapes, with a difference being the tape was two inches wide. The first recordings I ever made were done on this medium.

These tapes were lost, but I did have a mix of the songs that I transferred to computer to preserve them as soon as I could. As the years passed, I redid some, rewrote others, but there was one song that I always felt had great performances from the musicians, but would have liked to sing it again and keep the original work. Some years back, there was a software program for karaoke that supposedly could remove a vocal from a finished recording but it didn’t really work that well, certainly not for professional use.

Then last year it was announced that a cassette tape of John Lennon playing the piano was used to create a new Beatles recording. Using AI, they were able to separate the vocal from the piano, and then use it as the basis for a new track. Here’s the story of that project:

https://www.youtube.com/watch?v=APJAQoSCwuA

The AI used for this is called “stem separation” with each instrument called a “stem,” I guess like stems on a plant. Weird name aside, the interesting thing is that it uses newly developed algorithms to “hear” the different sounds and recognize them as drums, guitar or whatever and so is able to create a separate file for each.

Then early this year I upgraded my decade-,old recording software and discovered that it included a stem separator module. I loaded up my old song and was delighted to hear that it separated all the instruments perfectly, and now I could redo the vocal as well as make any other changes to the song; add another guitar or whatever. As a side note, this has become somewhat nostalgic for me, sort of like time travel to revisit old friends and play music with them, as well as my former self. Kind of trippy. 

There are concerns that AI will profoundly disrupt the economy, and just like any new paradigm or invention, this is to be expected. There are innumerable applications for AI, and they are coming regardless, and there will be unintended consequences that require both regulation and adaptation. My experience using this tool for my art is one very useful and rewarding example of how it will enhance our lives, and one hopes the value of AI will be greater than any turmoil it creates in this ever-changing world.

[Note: To get daily delivery of this blog to your email go to: https://petersage.substack.com. Subscribe.  Don't pay. The blog is free and always will be.]