Sign In

Communications of the ACM

ACM News

Give Me That AI Rock and Roll

View as: Print Mobile App Share:
A represention of artificial intelligence as a maker of music.

Software incorporating artificial intelligence is helping novices to compose music.


Music created by artificial intelligence (AI) has a long way to go before it will worry the likes Beyonce or Ed Sheeran. Still, some of its earliest incarnations are offering novices the ability to compose music.

Using tools like Orb Composer, Amper, and Jukedeck, people with no musical training at all can have AI help them play with sounds and create tunes in a matter of seconds.

"There's no reason to deny the fun of music to anyone," says Nick Collins, associate professor of composition at Durham University in the U.K., who has created his own music using AI.

Even better, the tunes generated by many of these AI tools are royalty-free.

In some cases, as with Amper, for example, people pay a small, one-time fee to use the song.  With others, like Orb, anything you produce is yours to keep, free of charge.

Says Jean-Pierre Briot, research director at France's National Center for Scientific Research (CNRS), "I think it's more constructive to talk about AI to help musicians to compose and produce good music," than to look to AI alone either for head-turning compositions or to help amateurs produce a magnum opus.

Even so, there have been some notable exceptions.

For example, YouTube sensation Taryn Southern used Amper as her primary composition tool to create the song "Break Free," which has received more than 1.6 million views on YouTube. With Amper, Southern was able to input her own melody, and then use the software's tools to select musical genre, mood, length, beat, instrumentation, and other variables.

As with many AI composers, Amper can generate a song in seconds, which then can be refined with the software's onboard tools. A raw production, for example, can be polished with the addition or subtraction of various instruments, or punched up with a quick change in tempo. Plus, for the complete beginner, the software can produce a tune even without the user having a melody in mind as a starting point, through the selection of genre, mood, and other variables.

Another notable creation of AI is "Beyond the Fence," the "world's first computer-composed musical," which was scored by an algorithm crafted by Durham's Collins with some help from Flow Composer, an AI composer from Sony Computer Science Laboratories created under the Flow Machines project of the European Research Council, according to Bob Sturm, a lecturer in digital media at the Queen Mary University of London School of Electronic Engineering and Computer Science.

AIVA is an AI music composer trained on the classical works of Mozart, Beethoven, and others.  The software, whose first album for piano and symphonic orchestra, "Genesis," was released in late 2016, takes a novel approach to AI music, rendering its creations in the form of sheet music, which are then brought to life by professional musicians. In June 2016, the Society of Authors, Composers and Publishers of Music (SACEM), an association of music composers and publishers in France and Luxembourg, officially registered AIVA as the first system of algorithmic composition to be recognized as a computer composer.

In the short term, expects the current wave of AI tools to unleash a massive amount of background music over the next few years. "The amount of AI-generated music may vastly exceed any music so far recorded because of the speed of generation," Collins says

AI music composers may inspire millions of music consumers to start creating their own songs. "Many amateur composers find it difficult to start; the blank page is intimidating with endless possibilities," says Oded Ben-Tal, a senior lecturer in music technology at Kingston University in the U.K. "The ability to generate (new music) at a click of a button could enable people to get started. Our folk-music-based AI (Folk RNN) suggests that this could be a tool in such a scenario."

Folk RNN is a website that gives users access to a recurrent neural network trained on transcriptions of folk music, in order to create a new folk tune.

Late last year, Sturm "played the role of a music creator" in a test of an AI composer to complete well-known Christmas carols. Said Sturm, "I found that the Folk RNN model facilitated my creative process in a variety of different ways."

Joe Dysart is an Internet speaker and business consultant based in Manhattan, NY, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account