composing · Uncategorized

Individuals/Companies Creating Computer Composed Music

(Note: I retitled this post on March 10, 2017.)

A topic that has fascinated me in recent years is that of computers writing music. They’ve actually been doing it for some time.

This is, of course, a controversial idea and questions abound. Will computer songwriters put working human composers out of work? Should we even call what computers compose “music”? Will the advent of computer penned music lead to wild, new forms of music? (Almost certainly.)

I’m not going to get into these debates here though I will link to this reddit conversation that touches on some of those questions. Instead I want to link to some examples of computer composed music and give people the lay of the land.

A few things to keep in mind. Much of this music is not only composed by computer but also “performed” by computer so it’s got that kind of quantized, somewhat robotic feel you might associate with EDM music. This might be a turn off (it is for me) and you have to make an effort to separate the melody and composition from the playing. (Even Bach sounds terrible when played badly.) Additionally, most of this music is not pop music—it’s not really designed to be catchy the way a pop song is. It’s more like “furniture music”—music meant to exist in the background. (I often use some of it to doze to.) Some of it is intended as soundtrack music for film, video and other media. Finally, some of it is written in the modern classical-style, which can be hard to digest (even when written by humans.)

So, with that series of caveats in place, let me present…

The Music of David Cope
Cope designed several computer systems that analyzed the work of existing human composers and then produced its own work. These compositions never quite set the world on fire but they’re certainly listenable. (Here’s the wiki on Cope.)

Melomics
This is probably my favorite computer composed music. As described here,  Melomics “applies non-conventional evolutionary algorithms to creating original music without human intervention.” Yeah… right. The following piece is a good example and you can find more on youtube.

Flow Machines: First AI Pop Song
Quoting from this article: “The first-ever pop song written by artificial intelligence was unveiled earlier this week by Sony CSL Research Laboratory. The song, which is called ‘Daddy’s Car’, was composed by an AI system called Flow Machines.” This is a case where the song was written by computer but then performed by humans who also wrote the lyrics. (So we’re still good for something.)

Eduardo Miranda’s Music
Miranda wrote the fascinating albeit math dense book, “Composing Music With Computers.” This piece in a decidedly modern classical style is described as an example of “computer-aided composition.” Not really my cup of tea, but still an achievement.

JukeDeck (Update Dec 23, 2016)
Jukedeck is, as I understand it, an AI music creation engine that has already composed half a million instrumental songs (mostly in a soundtrack style). It also enables users to create their own songs by letting them set things like, genre, tempo, mood etc. The songs aren’t flawless (they’ve got that robotic, midi feel), but impressive enough. I suspect this sort thing is already competing with human composed, royalty-free digital music (such as what I do) partly because it’s so damn cheap. With improved quality things can only get better/worse.

This video is a sample of music composed by Jukedeck. To really get a feel for it you should create an account and play with it.

Here’s an interesting article dicussing Jukedeck and AI music in general.

Google Magenta (Update Dec 24, 2016)
Nobody has done more with AI than Google and Magenta is their attempt to get AI creating music (and visual art.) Their first attempt, presented in a video linked here, doesn’t blow me away. But, hey… it’s Google. Let’s see what they’re doing in six months.

Amper Music (Update Dec 24, 2016)
Amper looks to be similar to Jukedeck (above) but more high end. Their product doesn’t appear to be live though you can hear their various sales spiels and some sample music over at their Facebook page. Worth keeping an eye on.

Update March 10, 2017: The Amper product is now in beta. I’ve played around with it a bit and it hasn’t blown me away. Some of the sound synthesis (e.g. the rendering of the actual timbres used) is klugey. Compositionally, it’s on par with Jukedeck. But let’s see what happens when the product goes live.

Brian Eno’s Scape (Update Dec 26, 2016)
Eno has been redefining the boundaries of music for decades. His Scape app, co-developed with Peter Chilvers, allows users to guide software in the process of creating music. This video provides an overview.

Aiva (Updated March 10, 2017)
Aiva is an AI software composer out of Luxembourgh. It (or “she”, to use the pronoun her creators use) is certainly impressive. While Jukedeck and Amper seem to mainly create chord vamps that repeat with variations, Aiva composes more advanced and classically based music. I do find the structure of her music a bit “off”—the dynamics seem confused—but give her some time.

Check out Aiva’s soundcloud for samples; the “Celtic Dance” stands out. (Don’t forget to “friend” her. She’s lonely.) Note that Aiva’s music is, as I understand it, played by humans (So we’re still useful. For a while.)

Popgun (Updated March 26, 2017)
This is an Austrailian company applying AI to various aspects of music including generation. Nothing to hear as of yet, but it’s worth keeping an eye on their site.

IBM’s Watson Music  (Updated April 20, 2017)
IBM’s AI Watson has been making headlines for awhile. This page list the various music focused projects Watson is tied in with.

“Push Button Bertha” (Updated March 26, 2017)
As a side note to all this, here’s an interesting article discussing an attempt in the 1950s at computer generated music. The main result was a Tin-Pan Alley style song called “Push Button Bertha” which can be heard at the link.

************************************

I will keep this article updated with links of note.

8 thoughts on “Individuals/Companies Creating Computer Composed Music

  1. This is a great topic of discussion! As it stands, there seems to be too much focus on the technology relating to the visual arts (VR, AR, etc.) and not enough attention on audio technology. The music produced by computers via programmers is a huge field to explore going forward.

    Personally, I have seen various ways to expand in this field within the music industry. As music industry professionals, we need to consider the technological resources available to us as this allows us to grow our understanding and push our careers forward.

    Thank you for your view and research on this exciting topic!

    1. Thanks! Yeah, there’s definitely a lot of interesting ideas that are possible. For years I’ve thought that a kind of developed music that went on for hours, slowly variating on its themes etc. would be quite interesting.

      I’d also be interested in a kind of AR for music. For example, you listen to music and get some sort of visual overlay that matches the melody (something like a laser light show.)

      1. It’s interesting that you mention that kind of music! I had a professor who composed a piece of music through a computer via algorithms that lasted exactly one year. The piece would modulate and change the year it was being played. I remember hearing the piece from time to time as I passed by his office.

        The AR for music is a neat idea. Perhaps there will be a want for that in the near future. Personally, I am attempting to use technology to compose the music I would like to hear. The examples you gave in this post will help me achieve this goal!

  2. My suspicion is that the most interesting computer written music will not be music that duplicates what humans can do, but focuses on what humans can’t do. I been reading a book called “Music, The Brain and Ecstasy” and he has a passage that gets into this.

    “… recent technology makes new varieties of musical device possible. Computers can interleave sounds in ways too intricate for performers to coordinate, and can continuously combine sounds of diverse pitch and intensity that no player could manage. Synthesizers can also meld (“morph”) sounds. The very idea of the discrete musical note could give way to more ethereal sonic entities.”

    I’d be interested in hearing what type of music you’d like to create.

Leave a comment