Live data from Hacker News

Generating Music (with an algorithm)

techblog.bozho.net

31–40 of 56 posts

Re: Generating Music (with an algorithm)

#33
post #25

I have generated quite a bit of music using evolutionary algorithms. My research is mostly about figuring out representations that are not at too low a level (eg 1 "gene" per note). I like representations that in some way represent the patterns, oscillations, and processes that seem to drive music forward over time. A paper: http://ncra.ucd.ie/papers/gecco2011_jmcdermott.pdf Some short mp3s: http://www.skynet.ie/~jmm…

Where did you get the instrument sounds?

Re: Generating Music (with an algorithm)

#34
post #22
post #3

Earlier quoted context omitted.

bad, or artificial? If "artificial" - that's the performance side, and it depends on the soundbank. If a real performer plays the same score, it would sound better. The point is to have nice compositions first :)

sounds are ok (to my untrained ear), and great effort building this! still, a little too random since it's algoritmic, maybe you can go for a little "coherence" throughout a piece, if it makes sense?

sure, any feedback is welcome. The thing is still in an experimental phase, and I'm trying to improve it bit by bit

Re: Generating Music (with an algorithm)

#35
post #28

Nice work. Is anybody aware of automatic music performing software? I think that this is similar to automatic generation of music but more structured and therefore easier to implement. I remember reading some studies about software performing classical compositions, but never found working prototype to play with.

I've read some papers about software used to help live performers, some sort of co-performer, or accompanying robot. But I can't find the paper right now..

Yap, there is a bunch of papers, but no publicly available prototype to try. Imagine having a python module:

from imaginary_music_module import piano_performer as p

p.play("chopin.ly", style="Pogorelich")

That would be nice :)

Re: Generating Music (with an algorithm)

#36
post #33
post #25

I have generated quite a bit of music using evolutionary algorithms. My research is mostly about figuring out representations that are not at too low a level (eg 1 "gene" per note). I like representations that in some way represent the patterns, oscillations, and processes that seem to drive music forward over time. A paper: http://ncra.ucd.ie/papers/gecco2011_jmcdermott.pdf Some short mp3s: http://www.skynet.ie/~jmm…

Where did you get the instrument sounds?

Garageband! It's a pretty crappy sequencer, but the synth sounds are fine. I imported the midi tracks saved by the algorithm and then mostly just chose from the preset synth sounds.

One problem I noticed sometimes was a limit to Garageband's polyphony in the pianos. I think it only allows 8 or 10 notes at a time in some cases (not sure exactly how it works) and my software has a habit of playing a lot more than that.

Post reply on HN