Swarm Music Album Black Allegheny from Evan Merz on Vimeo.

We’ve heard albums made by singular compositional minds and by bands. What would an album sound like if composed by swarm intelligence, by computer evolutionary models of individual agents or bots? That’s the question asked by composer Evan Merz in his new, full-length album “Black Allegheny.” (At top: the composer explains in a video.)

Western musical and creative tradition is steeped in linearity, from the forward motion of the music staff to the mythos of Aristotle’s Poetics.

So, maybe it’s little wonder that generative music – music that may not have linearity, or a beginning, middle, and end – hasn’t exactly been a big hit with the kids. Pioneers like Brian Eno have helped spread the gospel of generative music, but apart from lots of interesting experiments, there hasn’t been a lot of actual musical content. If you were to make a stack of generative music albums, your listening list would be fairly short.

All of that could be about to change. Programming code, the essential medium in which such models can be developed, is more accessible than ever. It’s also more visual, thanks to the popularization of tools like Processing, which can help make the abstract rules of generative music easier to grok. Merz, for his part, has taken on the challenge with his own Java-based software.

Saying the bots “compose” the music may be a little misleading. Generative music needs rules to operate. Before Eno, there was John Cage, whose “chance” compositions were as much defined by choices of materials as by ranges of indeterminacy. Merz makes a nod to Cage’s notion of a “gamut,” a collection of raw musical elements used as the input in the chance system. Here, though, Merz is aided by something Cage didn’t have: a swarm of intelligent “agents” can navigate those materials via simple rules, giving the music form and substance. Because they aren’t aware of the big picture, the music evolves more naturally, rather than being subjected to an over-arching narrative.

Or, as Merz puts it, “the tiny ant on the ground knows only what it sees around it.”

So, that’s the theory — what does the music sound like? Far from “ennui,” as Merz puts it, to me the results are organic. The structure is emergent from its materials, sounding almost like a natural physical process, like watching ice melt. The content ranges based on the gamut; like a lot of generative music, some sounds a whole lot like Brian Eno’s work. Others borrow from minimalist composers (Reich’s music itself might be seen as partially generative), and others take on an edgy urgency. The models that determine the bots are based on a popular, simple mathematical predator/food model, one often used in these works. Sometimes, you might imagine that evolutionary struggle playing out in the music.

You can read more about the process of developing this tool and the compositional ideas behind it at Evan’s blog:
Black Allegheny, Swarm Generated Music [Computer Music Blog]

For more explorations of sound and composition, check out Noise for Airports, which recently featured the work:

And you can stream the album or buy it for yourself for the light price of US$5 — though I’d like to see a software release, since that would mean each playback could be different. (Eno released an album in software form in the 90s, though tracking down the software now is evidently impossible – anyone with tips?)
Black Allegheny @ Bandcamp [Stream / download purchase]

<a href="http://evanxmerz.bandcamp.com/album/black-allegheny">Imperceptible Time by Evan X. Merz</a>

Swarm Controlled Sampler – Becoming Live from Evan Merz on Vimeo.

  • Man.. something versatile like this for M4L would be massive.. I've used a few Game Of Life systems before.. but nothing that I've been able to coax decent movements from..

    Anyone know of more generative tools like this for M4L or even just Max?

  • TweakingKnobs

    Freaking cool ,

    i also would love to see algorithmic music that Does have a beggining , middle and end.

    since this way it can tell a story.

    anyway this is still great.

  • @TweakingKnobs: well, there's actually no reason you can't use generative rules to do that, if you like. You choose the rules, you choose the materials.

  • Silent Procession

    I'm by no means an expert in this field, but it bugs me that it's presented as some kind of novelty, while swarm music is nothing new. I'd like to see credit where credit is due tbh!

    I don't know if Tim Blackwell invented it but he's done this in more elaborate ways since ten years ago.


  • Well, like I say in the article, it's not novelty that's been missing from generative music — it's quantity and variety. This particular algorithm is, again as I say in the post, fairly common. But a lot of times this sort of work has been relegated to technical demos or papers. There hasn't been all that much released in album form – not if you compare to, say, the amount of music made in a step sequencer. Repeatability is important to any technique. The fact that such algorithms have been so readily available and documented means to me that it's, honestly, a bit surprising that there hasn't been more musical output resulting from the work.

    I'm just trying to imagine how music history would have gone if the test were consistently novelty. "Species counterpoint? Dude, that's so early 17th Century. It's 1650, man."

    (There is certainly a sense in the video that novelty is the main emphasis, no argument there, but then that only further demonstrates why we need more context and continuity.)

  • neutral

    The imagination goes wild in relation to this fractal play room.


  • Dude, Bieber's been making music for swarms for like, a year already.

  • @TweakingKnobs: of course they can be born, grow and then die. So the generated sound can die with them.
    Maybe too sad; you cna just have a clock that acts as a conductor.

  • The point is the music sounds really nice and interesting. And the process is interesting too.

    Just a brief lookback:
    Fractal Music (Atari ST) – I'm user #52!
    Koan (now Noatikl)
    SoftStep (now ArtWonk)

    Sure Koan still works on any old machine, just let's look for Eno's album on the web!

  • "Gamut" is so appropriate:


    "[Middle English, the musical scale, from Medieval Latin gamma ut, low G : gamma, lowest note of the medieval scale (from Greek, gamma; see gamma) + ut, first note of the lowest hexachord (after ut, first word in a Latin hymn to Saint John the Baptist, the initial syllables of successive lines of which were sung to the notes of an ascending scale CDEFGA: Ut queant laxis resonare fibris Mira gestorum famuli tuorum, Solve polluti labii reatum, Sancte Iohannes).]"

  • mat

    Hey, I do love algorithm music too.

    Made a couple of songs with MIDImage (http://www.genjerdan.com/), here is an example: http://www.tonvibration.de/mp3/algorhythmik.mp3

    MIDImage makes music (midifiles) out of pictures. I just chose the instruments and cut some parts…not really arrange.

  • I understand your point about a lack of context and continuity – but for the lack of releases I think we can in part attribute this to the context that this music is often conceived for and also often best suited to – live performance/diffusion.

    In many generative systems I would imagine that the same compositional rules and choices provide varying results for each rendering – if we think of the composition as being more an evolving environment informed/influenced rather than dictated by choices, perhaps for some the concept of recording only one rendering of the environment may be at odds with the concept behind the work in the first place.

    I think this especially true when the system responds not only to compositional choices made in advance, but also to external input to the system during the performance.

  • I cannot find Eno's Generative Music I as data, but it is released as an album. That's agood idea, use the changing versions for live acts, and then choose, cut and edit an author's release.

  • Several of Eno's Generative Music I pieces appear in recorded form in the Koan retrospective "Dark Symphony"


    Links on the vimeo site also point to URLs where you can download the movies with a 5:1 soundtrack (recommended!).

  • thanks Umcorps!

  • tim

    You asked about tips on where to find Brian Eno's "Generative Music 1 (with SSEYO Koan software)" (see also: SSEYO Koan for more on Koan)?

    Sadly, GM1 is no longer available and has not been so for a long time. There was only one production run ever made (1996 – somewhere between 1k and 2k units) which sold out fairly quickly. It is likely safely in collectors hands and the only only place you might find it now would be, I guess, on Ebay or somesuch. Audio recordings of the generative music were never officially released and it required and relied on a particular Creative Labs soundcard to sound as intended.

    That said, a clean-room developed version of the Koan Generative Music engine that was at the core of GM1 is in active development, so people can still experience and create content that behaves the same. This development has the brand name of Intermorphic Noatikl. Noatikl was created by the developers of Koan, and it can play much of the old Koan generative content but is also evolving and has discarded the soundcard dependency. The Noatikl engine has been incorporated into Intermorphic Mixtikl, a 12 track generative music and loop mixing system for mobile (iPhone, iPad, iPod touch, Android [soon!] and later Symbian, MeeGo and Antix Game Player), desktop (Mac and Windows [standalone] with plugins [VST, AU and web browser]). Mixtikl contains its own DLS MIDI wavetable and modular synth as well as FX units, so the generative music it creates is portable across its system variants.

  • tim

    Thanks Mark

  • Thanks for the post, Peter.

    @TweakingKnobs: I didn't have the time to mention all of the other artists who have used Swarm Intelligence … but yeah, I am aware of them, and here's a short list: Tim Blackwell, Lawrence Bisig, Eduardo Reck Miranda and Tatsuo Unemi. As well as others who haven't published on the subject.

    None of them have ever released an album of just swarm generated music, as far as I know.

    I have a paper and conference presentation in the works, actually, which addresses a lot of the things that were brought up in this post and in the comments … I should get around to finishing it!

  • Random Chance

    Would be interesting to know how the software actually interprets a given configuration of the system in terms of different voices, pitches, durations, etc. That is the hard part, not the use of a particular swarn algorithm to generate a random walk of the search space.

  • anechoic

    In the 90's I composed some background music for a video game called Obsidian using Koan Pro…I also used it for a couple of ambient music performances…
    now I use the Schillinger System for my structures and some rhythmic permutations…music and math are the same thing

  • anechoic
  • Pingback: computermusicblog.com » Blog Archive » Black Allegheny, Swarm Generated Music()

  • Pingback: Música Criada por Algoritmos e Bots()

  • Simon

    The concept is cool and everything, but I've yet to hear generative music that doesn't sound like just a boring set of notes struck at random times and random (within a specified scale) pitch. Which it pretty much is, when you think of it.

  • Simon

    I just feel that the only thing interesting in these works is how the music is made. The output is most of the time a pile of dung.

  • Alex

    I agree with some aspects of what you're saying. If you just hit record and let these programs play on indefinitely, the end result will most likely sound too random. However, if you salvage the best bits from whatever the program spits out, there is sometimes usable material that can be looped, processed further, added to, etc.

  • Simon

    I agree it can be a pretty nice sampling source.

  • It would be nice to have this in M4L. That way it would be possible to put the M4L swarm instrument on a channel and record the output to keep the pieces that I like and use those as building blocks for a new song. Where can I find those swarm algorithms? Only one more question: if someone uses swarm algorithms to make a song, who is the composer? The algorithm or the person who wrote the algorithm?

  • @tim: yes, we know noatikl and the rest.
    Someone should sign a deal with Eno to release GM1 (or even better GM2, 3, 4…) for this system!

  • Pingback: Connect the Bots: Black Allegheny, An Entire Album Made by Algorithmic Swarms | VJ Heaven()

  • Miller Peterson

    I think Random Chance has got it right – the crucial element would seem to me to be the mapping from the low-level swarm algorithm into musical structures.

    Also, it would be interesting to hear why these patterns are supposed to be interesting. It's hard to discern any link between what's going on in the visuals and the music. Is there something about the way the music *sounds* that is unique?

  • Pingback: swarm music – black allegheny – abulafia()

  • Tracey Holliday

    I need to hear just what will do with that!?!

    Davidoff 4000