April 14, 2012
Las Tortugas Comen Queso...
by Chris Randall
I'm currently writing my next EP, and this album is interesting inasmuch as the particular set of constraints I've set for myself this time involve using only generative techniques for the sequencing. That is to say that while I'll be recording to and mixing from Live, there are no Live clips playing MIDI; all the hardware and softsynths are driven from generative methods created in Max4Live.
Simultaneously, I decided I wanted to finally learn Spanish, so I joined Duolingo, which is a very nice service. However, many (most? all?) of the learning phrases are generated algorithmically, and I'm left with the lingering impression that there is some synchronicity here. Or, in other words, the turtle eats cheese.
Now, you and I know that turtles don't eat cheese. Or if they do, they probably regret it later. However, the algorithm that generates statements for me to translate in Duolingo doesn't know that. It just does animal + verb + food. Thus, we get: las tortugas comen queso. On the one hand, the sentence makes perfect sense. If you actually saw some turtles eating cheese, you'd turn to your monoglot friend Javier and say "las tortugas comen queso" and he would no doubt be in full agreement.
On the other, you're never going to do that. Because you most likely will live your whole life without ever seeing a turtle eat cheese. So the sentence itself, when used as an example of language, is correct, but ridiculous. This is largely the same thing with generative music, I discover. I was having a conversation with Matthew Davidson (stretta) about this via email earlier this week, and he pointed out that generative techniques probably best serve when they are just one tool in the musician's toolbox. I, ever the purist, bristled a bit at this; when I set myself a paradigm for recording, I don't do it by half measures, and the tracks I'm creating play by themselves, with zero input from me, and the resulting recording is already pre-mixed, even.
However, this methodology of mine, while crafty, usually results in a cheese-eating turtle. That is to say that my algorithms are choosing notes that are harmonically correct, but not necessarily something a person would play. I suppose this is the point of the exercise, but I was hoping for a more organic sound.
So my question is thus: what are your feelings vis-a-vis generative techniques? Are you like me, where it should be an all-or-nothing affair? Or is some subtle randomization to a hi-hat pattern or hand-editing MIDI output more to your taste? What's some good electronic music that was created wholly generatively?
Simultaneously, I decided I wanted to finally learn Spanish, so I joined Duolingo, which is a very nice service. However, many (most? all?) of the learning phrases are generated algorithmically, and I'm left with the lingering impression that there is some synchronicity here. Or, in other words, the turtle eats cheese.
Now, you and I know that turtles don't eat cheese. Or if they do, they probably regret it later. However, the algorithm that generates statements for me to translate in Duolingo doesn't know that. It just does animal + verb + food. Thus, we get: las tortugas comen queso. On the one hand, the sentence makes perfect sense. If you actually saw some turtles eating cheese, you'd turn to your monoglot friend Javier and say "las tortugas comen queso" and he would no doubt be in full agreement.
On the other, you're never going to do that. Because you most likely will live your whole life without ever seeing a turtle eat cheese. So the sentence itself, when used as an example of language, is correct, but ridiculous. This is largely the same thing with generative music, I discover. I was having a conversation with Matthew Davidson (stretta) about this via email earlier this week, and he pointed out that generative techniques probably best serve when they are just one tool in the musician's toolbox. I, ever the purist, bristled a bit at this; when I set myself a paradigm for recording, I don't do it by half measures, and the tracks I'm creating play by themselves, with zero input from me, and the resulting recording is already pre-mixed, even.
However, this methodology of mine, while crafty, usually results in a cheese-eating turtle. That is to say that my algorithms are choosing notes that are harmonically correct, but not necessarily something a person would play. I suppose this is the point of the exercise, but I was hoping for a more organic sound.
So my question is thus: what are your feelings vis-a-vis generative techniques? Are you like me, where it should be an all-or-nothing affair? Or is some subtle randomization to a hi-hat pattern or hand-editing MIDI output more to your taste? What's some good electronic music that was created wholly generatively?
54 comments:
I'm sure there are examples of good electronic music made in this way, but I'm drawing a blank. I view generative music as more preliminary stuff to slice up and play around with, but then that's my general approach to everything.
There is a flatness to it I often find, that it needs something from outside to make it more that just an interesting but one-dimensional demonstration of technology.
There is a flatness to it I often find, that it needs something from outside to make it more that just an interesting but one-dimensional demonstration of technology.
Hit random until it sounds good !
I cant remember the last time I actually wrote an acid bassline since I got d-16 Phosycon.
I cant remember the last time I actually wrote an acid bassline since I got d-16 Phosycon.
i shape the generative stuff to be more "me". either capturing the midi data and editing modifying etc or just using it as inspiration.
there are chunks of autechre tracks that to me sound like they came out of an algorithm or were inspired by jam sessions with algorithms. there's stuff on LP5 that has that kind of arc.. starts at A and ends at Z and the progression is just some kind of slope with some hills and valleys to get there.
i recall messing around with oblivion midi generator on windows years ago and getting some notes and sequences that led to a forehead slap as they just shadowed bits of LP5 in very close ways. that being said.. the layering, sound design, emotion and overall nail on the head factor of that record is staggering to me no matter how it was done
there are chunks of autechre tracks that to me sound like they came out of an algorithm or were inspired by jam sessions with algorithms. there's stuff on LP5 that has that kind of arc.. starts at A and ends at Z and the progression is just some kind of slope with some hills and valleys to get there.
i recall messing around with oblivion midi generator on windows years ago and getting some notes and sequences that led to a forehead slap as they just shadowed bits of LP5 in very close ways. that being said.. the layering, sound design, emotion and overall nail on the head factor of that record is staggering to me no matter how it was done
In general I think you can get good interesting generative music (except for drums maybe) for slow and ambient pieces (40-50BPM) . I use Intermorphic Noatikl and feed it into 3-5 synths (real and VST) and I'm never (well mostly never) disappointed, it also allows you to focus exclusively on the realtime interaction of the synth sounds and timbres and not be bothered with keyboard virtuosity (of which i possess none) and it also dumps its MIDI stream into a file that you can always go back and retrieve it . The trick is to go after those rules tables in the program pretty hard to keep it from just being totally random.....Lately I've been using Julien Bayles' Designethemedia's M4Live tools, Matrix, Protoclidean, and Granulizator which are all probability based and a lot of fun, especially if you get a M4Live LFO tool to twist the dials for you...
Is AD Replicant generative? What about BigSeq and Tattoo random resets. They're certainly a touchless audio carwash and add interest to otherwise repetitive beat tracks.
What if you just turn a modular loose with a lot of free running CV cycles patched in here and there....generative sounds come out for sure but when does it cross into the threshold of being music and no longer a noise piece? Seems like that's what everybody with a Buchla ever does.
As for good generative music, you like Brian Eno or you don't but he's got a lot.
Is AD Replicant generative? What about BigSeq and Tattoo random resets. They're certainly a touchless audio carwash and add interest to otherwise repetitive beat tracks.
What if you just turn a modular loose with a lot of free running CV cycles patched in here and there....generative sounds come out for sure but when does it cross into the threshold of being music and no longer a noise piece? Seems like that's what everybody with a Buchla ever does.
As for good generative music, you like Brian Eno or you don't but he's got a lot.
PS ...btw I think that Los Quesos Comen Tortugas is probably more fun
Also my favorite Spanish alliteration? "como como como" which means "I eat like I eat"
Also my favorite Spanish alliteration? "como como como" which means "I eat like I eat"
My submission for the AI comp "Heterodoxic" had a lot of generated basslines in it. (Since it was not chosen, we can assume it was cheese eating.) I had two interlocking 4 bar basslines and ran them through Intermophic's Noatikl. It can mutate patterns endlessly and produce a MIDI file containing its output. I listened back to more than an hour of generated output and chose the best bits that became the bass for the verse, chorus, middle eight, etc. I played or programmed all the other parts.
Personally, I would never attempt to generate an entire piece because pure generative music sounds random and pointless. I use it to get unstuck -- to come up with something fresh that I wouldn't normally play.
Personally, I would never attempt to generate an entire piece because pure generative music sounds random and pointless. I use it to get unstuck -- to come up with something fresh that I wouldn't normally play.
i don't know that generative and random are synonymous. generative can be a set of rules controlling something. those rules can be things that are intensely programmed to generate things within specific parameters as part of a larger set of rules.. nests within nest of things that have a starting point and rules to shape their progression over time while still being generative.
i don't think that adds up to 'random' or 'pointless sounding'.
i think there are different depths of generative.. hitting the random button on a sequencer or setting parameters on one part of something to be generative and then there is the process of controlling na entire studio with a set of algorithms and 'seeds' for those and have ways built in for nudging the process at various times to you can continually have some effect on the process and make it turn left/right etc.
i don't think that adds up to 'random' or 'pointless sounding'.
i think there are different depths of generative.. hitting the random button on a sequencer or setting parameters on one part of something to be generative and then there is the process of controlling na entire studio with a set of algorithms and 'seeds' for those and have ways built in for nudging the process at various times to you can continually have some effect on the process and make it turn left/right etc.
I don't like completely auto generated music so I prefer working with Axon. It's not random but when 3 neurons are activated and triggering each other it's already out of my control. The output is useful as is but usually I end up editing the MIDI clips and cut and paste until I'm satisfied. I don't use the built-in synth so much and connect it to other VSTi's or hardware instead. I don't mind blending the results from Axon with non-generative parts (and that's how I use it) but then again, it's not totally random.
I'm with boobs (and by extension Trackdriver.) I am most certainly not talking about just throwing randomization at things. Axon is a good example of what I mean, as it is essentially a VST version of a Pd sketch I made to do exactly the same thing. It isn't random at all, being entirely determinative. However, the patterns are not something we would normally come up with.
-CR
-CR
I like incorporating generative elements but I also appreciate being able to exercise some level of control over them as they burble away. I use a Monome clone that I built for myself, and one of my favorite apps is richochet (link [post.monome.org]) which is generative, but allows you to nudge the ball, so to speak, and dynamically influence the app as it does its thing. If you leave it alone it will happily generate data, but there are certain parameters you can "play" during usage.
Comment:
Sorry, commenting is closed for this blog entry.
