Generative systems

[kisielk]

Awesome this is why I love this forum. Learning new things every day…

[kisielk]

Also I’m increasingly jealous of all the people who spent years studying and playing music in school as opposed to something boring like electrical engineering…

[glia]

i didnt even know what ee was

now that i do it wouldve been a great foundation

[jasonw22]

I couldn’t afford college. Good thing I studied programming and graphic design as a kid. In later life such interests have made my music hobby affordable.

[hecanjog]

The grass is always greener I suppose – I’ve felt like I’ve been trying to play catchup with math and computer science fundamentals my entire adult life and I feel a bit of the same jealousy towards friends with math/compsci/ee backgrounds.

[sellanraa]

As a history teacher, I like that all of my interest in the more detailed mathematical elements of music and MAX programming ends up feeling like some sort of medieval alchemy grafted into 21st century living

[beo]

i would call a generative system a closed structure which is able to give an output.
in other (negative) words : a structure that doesn’t need an input to give an output.

for instance see attached file. it’s an aalto patch. simple and scanty
means.
Untitled.zip (1.7 KB)

[glia]

i waited and am surprised that nobody has referenced laurie spiegel yet

[guy]

I highly recommend this paper by Philip Galanter on generative art: http://www.philipgalanter.com/downloads/ga2003_paper.pdf
The focus is on generative art more broadly, mainly visual rather than music, but most of the discussion applies to music too. Galanter provides a useful and succinct definition of generative art:

Generative art refers to any art practice where the artist uses a system, such as a set of natural language rules, a computer program, a machine, or other procedural invention, which is set into motion with some degree of autonomy contributing to or resulting in a completed work of art.

The key point is that the generative artist cedes (gives away) some amount of creativity or creative control to the generative system. I find this definition useful because it doesn’t rely on specifying processes or technologies - it’s broad enough to cover generative music and clear enough to draw a line with other categories of practice (e.g. computer music).

[laborcamp]

I mostly agree with this.
Though the idea that the outcome is a “completed work of art” is somewhat problematic to me. I think the generative quality introduces an element of openness to the work. At least I always look for strategies that assure an aspect of incompleteness, something that is happening, rather than something that has happened.

[Marco_Raaphorst]

Here’s my first track using generative techniques:
https://soundcloud.com/raaphorst/evening-lights-alt-5

In Propellerhead Reason 9.1 I have created a huge generative rack full of devices which are triggered randomly. It sweeps between randomised pentatonic notes. The chords are also generated randomly based on these notes.

Sound is not being randomised, only the melody lines and chords. For a new experiment maybe I would reverse the proces, keep the notes and chords the same but change the sound randomly. It’s interesting because this track is 22 minutes long but it could also be 59, 3 hours or 1 minute

[kisielk]

This is a really cool set of interactive slides on generative music: https://teropa.info/loop/#/generative

[dan_derks]

Everything about that person is excellent for this thread. They just released a browser version of Laurie Spiegel’s Music Mouse that is utterly fantastic. MIDI out, as well. Playing with it for the last few days has taught me so much about her approaches to rhythm through melody.

[kisielk]

Oh cool, I didn’t even notice that was made by the same person.

Edit: Actually, their magic mouse Javascript is in this presentation! (haven’t made it all the way through)

[Pampalini]

There’s also generative in the sense of generative grammar where you define a set of rules like

S -> NP VP
VP -> NP V
...

together with their corresponding probabilities and then start sampling from it recursively. Think of a Markov chain with some recursion thrown in. It’s a fun way of making music by only defining its grammar.

[kisielk]

Definitely There’s a good demo of that in the link I posted!

[Pampalini]

That was a hell of a set of slides. Very engaging.

[simondemeule]

A question to the geekier crowd of lines: what are some interesting computational methods you know for generating music? And conversely, what are some interesting models for analysis?

In a theoretical computer science course I’m taking it occurred to me that formal grammars could be an interesting way to generate and analyse music (although mostly generate, as I suspect machine learning is better suited for analysis) — and it just happens that people have indeed explored that. I stumbled upon this paper a few days ago — Grammar Based Music Composition — and I’ve been wondering if some live coding environments implemented similar things.

I’m looking forward to hearing your thoughts on this!

[charlieroberts]

In the live coding world, Renick Bell is well-known for using L-systems in his work, using his Conductive environment. I also read this blog entry a couple weeks back on L-systems / SuperCollider / Live Coding: https://theseanco.github.io/howto_co34pt_liveCode/3-6-L-Systems-For-Rhythm/

[alanza]

I don’t know basically anything about this at all, but within formal grammars, your post makes me wonder about using finite state automata to, like, generate progressions or something.

There’s this concept, following Schoenberg and others, of musical set theory which, to be honest, I’ve always kind of disdained—not for good reasons! just because it struck me as barely scratching the surface as far as realizing mathematical concepts formally within music.

Anyway, one way to think about musical set theory is fixing a set (classically, a collection of “pitch classes”, i.e. note names), and then exploring some permutations of the set. This reminds me of mathematical group theory, where often you are concerned with studying collections of permutations or symmetries.

One interesting wrinkle this viewpoint suggests to me is restricting or changing the kinds of operations you can perform on your set, so that rather than just Schoenberg’s transposition, inversion and retrograde, maybe you could let yourself swap the position of a few notes in a melody, or something.

Even studying symmetries of a small number of objects provides us with plenty of interesting algebraic things to think about, so surely there must be a musically interesting way of interpreting that.

---

Of course there’s also that paper from whence we get the term “Euclidean Rhythm”, which basically comes down to thinking of rhythm as trying to solve some division problem (put 3 beats inside a group of 8 beats) with constraints (space the beats as evenly as possible). Maybe too general to be fruitful, but I suspect interesting things are to be (re-)found if we either change the problem (say, make a 7-note scale from 12 notes), or change the constraints (emphasize beats away from the “four on the floor”)