THE LOSS FOUNDATION BLOG

personalized expeditions in music from a musician who doesn't matter

Unicorns revisited

Posted by r on February 28, 2012

(Edit cleanup 3.19.13: This post is on an elaborate if totally silly chance-based system of composition that I came up with but have only employed even partially for one half-song, and have been thinking about using again at greater length for five or six years now, but I definitely haven’t gotten around to it yet. Recommended for only the toughest of tabletop RPGers and/or system-loving noisehounds.)

Last week the bassist I’ve been jamming with for the last few months brought in a new-to-me guitarist, because I’ve been asking for one. Next up– since I seem to be stuck on guitar in this project so far– is a drummer, and hopefully– if I get my way– a female vocalist who does nothing but screech on a glass-breaking order. (Wes, the bassist, is sort of fighting me on this one, no matter how many times I tell him he really needs to listen to more Pre and Diamanda Galas records until he enjoys the idea.)

Anyway, I am liking the vibe the three of us currently have, and I am hoping that our new guitarist friend will stick around because I think it’s starting to look promising. One problem Wes and I have consistently had, though, and that now the three of us have collectively, is how to start writing.

Writing has always been really personal for me and it progresses pretty naturally as I work alone, any white space between useful A-HA moments being inconsequential as I’m the only annoyed party. Throw another person in the room and things become Awkward. It’s also hard to get things to move beyond simple one-riff jam sessions.

I proposed to Wes, who is a coder, that we really ought to make a mini-program called The Decider (replete with background images of Dubya in finest form). The Decider would arbitrarily pick out some basic parameters as starting points for collective jamming– key, meter (or sequence of meters), tempo, possible modality or combination of modes, maybe even a rhythmic cell or a couple of notes to use as a point of departure, and a couple of adjectives to reflect upon in beginning.

Then The Decider would count down five or seven minutes or so, and if you didn’t have something good stemming from those parameters by then, it would move on to the next set of starting parameters.

The Decider, then, would just be the equivalent of a musical icebreaker to stop everyone from plugging in, then sitting around and scratching their heads. It is hardly algorithmic composition– a topic I’ve always been interested in, and would always like to know a lot more about. (I have this book currently sitting on the coffee table, courtesy the library. It’s totally fascinating stuff, dense as all hell with mindblowing trivia, and I blushingly admit that I’m only really about 30 pages in, and actually he hasn’t even talked about music yet. I would buy it happily in hardback if the only remaining copies in that format weren’t going on Amazon for $1,300 for some completely stupid reason.)

Last night, I also accidentally stumbled upon an old French literary movement called OuLiPo (or Oulipo, depending on who you ask). These dudes are / were also interested in creating restrictive systems and rules to generate literature and poetry. Hardly new; the ancient fixed-form poem (sonnet, sestina, etc.) is a core source of inspiration for the types of things these dudes were up to. They just took the concept of a priori contraint a lot farther, or at least some of them did.

The concept behind quintessential-Oulipian Queneau’s Hundred Thousand Billion Poems is especially astounding. It’s easy to contemplate a set of musical works that could replicate Queneau’s practically infinite splice-and-dice reconfigurability. But without the semantics required in poetry, said set of musical works would be much less of a head-exploding achievement.

Tying this all together, the other day I re-listened to the two minutes of music representing the proof of concept for a fake band and real songwriting method I was calling Unicorns Are Real and They Can Also Be Vampires. And wow, it’s still heavy as hell, and I’m beginning to think it’s time to revisit this concept.

At the time, I was calling the UARATCABV concept “Excel rock.” This is a misnomer, because Excel really isn’t required; you just need a system of random determination. Like The Decider, but to a much more specific and directed degree, the “roadmap” for UARATCABV songs mainly determines the entire framework of the song in advance. The equivalent of dice rolls determine everything from the form of the song to the (rough) order of pitches found from moment to moment.

There’s no reason why, again, you couldn’t create a single app that would generate all the materials laid out in my “2.0 roadmap” and start the process of writing a UARATCABV song. Anyone want to help me code?

Said “roadmap” straight copied-and-pasted from a 2007 Livejournal entry follows after the jump.

1) Run random.org to generate a single integer between 3 and 7. This is a really important number – it is the number of distinctly identifiable sections the song will have, akin to a verse or a chorus section. Each section is still based upon a core riff, but we’ll get to this later; just be aware that I’ll probably start referring to “riffs” instead of “sections;” each section is in fact entirely comprised at the core of a single riff, aka a short chunk of repeating material that I’ll be playing with further. This number of distinctly identifiable sections is heretofore referred to as variable s.

2) Do two more random.org integer runs (repetitions allowed). The first run produces a single integer between s and (s+4); this is the total number of sections in the entire song, which I will call variable ns. The second run then produces a string of numbers of length ns with range 1-s.

What this all decides is the actual “letter layout” of the song, what sections come in what order and when they are repeated. A supposedly traditional rock song, verse-chorus-verse-chorus-bridge-chorus,might look like this when translated to letter form:
ABABCB
Its ns would be a value of 6, but there are only three distinctly identifiable sections (in other words, s = 3). C, the bridge, is never repeated, but A (verse) and B (chorus) definitely are. It is possible that random.org would spit back these exact values (s=3, ns=6, run 2.5 producing 121232 = ABABCB) and I’d then end up with a traditional rock song. However, even with the kind of contraints I’m putting on the system, this is really pretty unlikely, so I’ll almost certainly end up with weirder-looking forms that I have to deal with and move on.

My “test” run gave me s=5 and ns=8 with run 2.5 providing “45413231” = the form ABACDEDC, which is a really nice-looking “compound” form with a form of closure at the end from whatever will become the C riff/section. Good stuff, trust me.

Where this is a little confusing is that the numbers from run 2.5 decide which riff/section becomes A, B, etc. Riffs 4 and 5 here will destined to become letters A and B respectively, because that’s the order in which they’re heard and that’s just the way this crap goes. Don’t get confused yet.

3) This part is also part of my UARATCABV “roadmap 1.0,” the same one I used for “Abominator,” and generates the length of each “initial riff” in measures, with a range of 4-10 measures total, for each “member” of s. If s=5, as it does here, then I have five integers to generate.
The test run gave me 4 9 7 6 10; riff 1 will be 4 measures long, riff 2 will be 9 measures long, and so on. (Again, this doesn’t correspond to the final “letter structure” of the song; riff 4 = A, so the “core riff” that comprises A is 6 measures in length.)

4) The next step was also part of roadmap 1.0. We’ve determined how long each riff will be, now we have to determine what meter (time signature) each measure of the riff will have. I run this process s times, since I have that many riffs to fill up.

The number of digits generated by random.org here is determined by the length of each riff in measures, while the range is always a number from 1-9. The numbers produced map onto the following arbitrarily-chosen meters:
1: 3/8
2: 3/4
3: 5/8
4: 6/8
5: 7/8
6: 4/4
7: 9/8
8: 5/4
9: 7/4

Riff 1 is 4 measures long and the test run from random.org gave me 4436, which means that this riff will be comprised of two measures of 6/8 time followed by a bar of 5/8 and a measure of 4/4, exactly in that order. I repeat this process until the metrical structure for all of the riffs is decided. This is crazy enough, but it gets crazier from here.

5) This stage determines how many times each riff will be repeated within its section appearance. Remember, we had ns=8 in the test run with a final formal outline of ABACDEDC. The number of integers generated by random.org = ns, while the range of the numbers is from 2-6.

What this means: When we have sections that repeat, as here with A, C and D all appearing twice each in the form, this means that they could (and probably will) be different lengths despite the fact that they are coming from the same riff. My test run called for 3 repetitions of the letter-A riff (riff 4) in the first A section, but 6 on the second run, while letter C will be 33% longer the first time it is heard (three repetitions of riff #1) than the second (which came up with only two repetitions).

This is cooler than shit, even though I am actually simplifying the rules a little bit (there are a few further rules to help deal with excessive repetitions of excessively long riffs that may have been made excessively long during steps 3 and 4, but let’s just forget all that for now). Still with me? OK.

6) OK, this is the absolute-madness stage. I’m not content with having ridiculous formal outlines, constantly changing meters in the riffs, and uneven numbers of riff repetitions, OH NO. No, from here there’s a further formula that decides “transformations” of each riff in each given letter-section.

The first two repetitions of each riff will always have the same metrical structure and pitch content. If step 5 called for the riff to be repeated more than twice, though, that means it will get TOO easy to follow and nod your head to, so here’s where the craziness kicks in.

I roll a “six-sided die” (random.org range 1-6) a number of times. How many? The answer is “the number of measures in the longest riff as decided by step 3, MINUS 2.” Why? It’s either obvious or it isn’t. Oh well. Here, my longest riff is 10 measures, so I have random.org “roll the die” 8 times and look at the results.

Each number on the “virtual die” requires me to something to each riff through each subsequent repetition past the second time through… or, it requires me to do nothing. In fact, if I roll a 1 or a 6, I do exactly that… nothing.

If I roll a two, it means I need to add / insert x number of measures after measure #y in the riff. If I roll a three, I need to do sort of the same thing, but REMOVE the measures of the existing riff there, instead of adding more in.

Rolling a four indicates that I will take measure x from the riff and repeat it y times before that repetition of the riff is completed. And rolling five means I’ll transpose x measures of the core riff starting after measure y by z semitones.

Where do x, y and z come from? They are separate die rolls, one for each riff transformation.

There are some other rules too, which force numbers certain ways in certain situations and again make sure that no matter how weird all this becomes, it will still be “stylistically correct.” I won’t get into them here.

In fact, in general, this stage is a huge pain in the ass to both execute and explain, but in a nutshell it means that no riff will ever be exactly the same more than twice through. It will still be similar enough to be connected audibly, but it’ll fuck up your would-be head-nod attempts something awful and simultaneously rock yo’ nutz even harder (yes, that’s right, ladies). It is a very cool trick, and it may be too ridiculous to actually pull off in performance (not that this shit is very easy to play as it stands). We’ll have to find out when I start trying to record songs with all this going on. All I know is Meshuggah seems to be capable of this kind of crap, but I am not as awesome as Meshuggah.

7) Finally, the tone row is generated in twelve sequential numbers 0-11 and a traditional twelve-tone matrix with 48 distinct row permutations is created from that. Nothing terrifically exciting here, just ohhhh generating all the pitch material for the song, move along.

8) Then I somehow write and presumably record the bloody thing.

 

Leave a Comment

Your email address will not be published.