Re: future of music essays

Subject: Re: future of music essays
From: Kevin Austin (
Date: Mon Apr 18 2005 - 22:02:25 EDT


Perhaps you are being too 'purist' about how other people may go
about the process(es) of creating "algorithmic" or 'modeled' pieces.
There may be many degrees of independence.

The composer may decide to provide a number of rules of composition
and allow the computer to chug on and create "in the style of". An
example could be the 'computer-generated' versions of (say) the first
16-18 bars of Satie's Gymnopedie. Such a "composition" could be
provided with a basic set of about 6 rules and some "knowledge" of
what a human hand can play, and produce some remarkably similar
sounding pieces.

They don't need to be indistinguishable from the style of Satie, just
adequately reminiscent. This isn't an attempt to write Bach from
scratch. And if one's style has no specific historical precedent,
then it may be possible to program a generator. I also have found
that it is useful to fiddle a bit with the results, because this
process is less about showing a computer can do it, than generating
blocks of successful data.

Let me give you a harmonic generative rule for the left hand of the Satie.
(1) Create two seventh chords which have 2 common tones
(2) alternate between them

A computer-composition purist would object, but it works to produce
the basis for acceptable accompaniments.

I am reminded of seeing someone trying to write a generalizable music
optical character recognition software. They were showing the
problems of determining how many lines were on a staff. I asked why
not simply enter this in a data field. The response was about the
"purity" of the program being able to work without human action. It
was not clear if they were trying to respond a HAL-like situation.

At 9:55 PM -0700 4/18/05, Eliot Handelman wrote:
>Ned Bouhalassa wrote:
>>I wrote formulaic, almost robot-made music, not algorithmic
>>composition. The form of many of top 40 pieces is identical, and
>>the chord progressions are from a tried-and-true mold, right?
>Maybe, but figuring out a robotic strategy for when you can do this
>or that in such a way that it works is still completely beyond
>anything anyone knows how to program.

This archive was generated by hypermail 2b27 : Sat Dec 22 2007 - 01:46:07 EST