[PD] pd-based procedural chord progression database.. / rhythm machine

[padawan12]

Wow such a lot of stimulating musical discussion
on the list this week. 

On Mon, 14 Aug 2006 11:24:56 +1200
Damian Stewart <damian at frey.co.nz> wrote:

> Chuckk Hubbard wrote:
> 
> > I have no idea how to make a patch that would insert pivot tones
> > somewhat randomly.
> 
> i am finding the concept of 'boredom' to be very, very programmatically 
> fruitful lately. presumably i know (or can find out) all of the notes that 
> are currently playing, or have been playing for the last little while. have 
> a boredom counter that increases faster when the changes are fewer. when 
> boredom reaches a particular threshold, pick a chord target and start 
> inserting pivot tones... increasing the frequency as the boredom approaches 
> the (higher) chord-switch threshold (of course, the new tones would 
> interact with the boredom counter - oh i do love feedback :-))...

If things start to get too chaotic or unstable at longer scales you could
always feed-forward, in other words not include your new pivot notes in
the analysis until the change occurs. 
> 
> hmm. i have a centralised rhythm machine at the moment that runs constantly 
> and bangs about 50 different sends at different periods and different 
> phases (n, 2n, 3n, 4n, 2n-then-3n-then-3n, 3-2-3n, 3-3-2n, etc). it's 
> attached. it's based on a book called 'Hearing in Time' by Justin London, 
> which is wonderful and has informed my ideas about techno no end.
> 
> so have a similar thing for assigning tones -- rather than each instrument 
> having its own tone input, instruments just listen to a master tone 
> assigner, that has a bunch of different sends (eg t-root, t1, t2, t3, t4) 
> for the tones in order of diminishing importance/salience to define a 
> particular harmonic context.

Sounds great. I've been playing with some similar ideas in composing
lately. The idea of departing from instruments strictly following a 
score and putting some "intelligence" into instruments or groups of 
instruments seems good. A while back I was trying to get a tune based
on a "rythmn section" and a 'brass section". The rythmn section was
the drummer, and the bass, with the bass following a "busyness" parameter
and a "dynamics" parameter. The problem is, generally with realtime
dataflows like puredata, that you have no look-ahead possibilty. 
Like the problem of predicting duration for tied notes when your score
is generative. Every function you use to derive a parameter is necessarily
following along behind what has already happened. 

I thought about it two ways. Either you want to keep a buffer of one or two
bars and fill that first, in which case you kind of defaulted back to
just following a dynamic score. Or you imbue your "players" with some kind
of anticipation. Say we are in bar N and bar N-3 was was a busy cadence and
bars N-2 ans N-1 were straight then assume we are playing in 4/4 and get ready
to end the bar conclusively. This a kind of musical autocorrelation.
I got into a bit of a mess programming that but I must have another try again
soon. Thanks for the multi phase bang box , will play with that later.

Andy

> f r e y
> live music with machines
> http://www.frey.co.nz
>