I'm really looking about insight into process here. Did you take a probabilistic approach? Sonifying mathematical algorithm, etc?
When I played around with it, I made a mental map of what I wanted to achieve. Since generative music is basically coding for me, I used the same approach when designing programs.
You got your basic black box layout:
The input in this case is a set of random (or pseudo-random) numbers. Choosing the random source is probably the most fun aspect, because you can get really funky with it: Radio static, space rays, fractals, the text of a book etc.
The output determines a few things about the black box: How structured should it be? Do I want 12-TET notes? Should it be slow or fast? And so forth.
For each synth in the array, the black box opens to this content:
I then decide on per-voice basis, what parameters I want to manipulate and which should stay static. This decides on the number of quantization functions I have to write for the instrument. Let's say:
There's a number of way to approach this, dependent on your aesthetic preferences. For example, for a droney background pad the function for triggers doesn't need to be that strict when it comes to hitting a tick. In the same example, you'd probably want a generally long envelope, so the functions need to be interdependent (maybe sharing the same function with inverse output).
For notes you'll generally want a quantization that translates to harmonic steps.
If I want to have a global structure, like arrangement, chord changes, speed changes etc, there'll have to be functions which control the other functions accordingly. The quantization functions for notes would need an input parameter which harmony they need to shift to in case of a chord change, for example.
Of course, there's always the Just-Try-Shit-Out approach in which you slam a random source to some parameters and see what happens, knob twiddling your way to success. I've spent hours just exploring synthesis that way.
Edit: Here's a very strict example. Arrangement is fixed, and the only random aspect is the probability of each note trigger.
Last edited by µB (May 1, 2012 9:56 am)