Homework 5:
A midi file contains events whose parameters generally come from any of three sources (possibly in combination), either direct from score, by hand editing, or from triggers capturing events in real-time performance. Score events translate to "perfect" mathematical proportions. Hand-edited events can be anything, but tend towards perfect as well, just because that's what midi sequencer palettes provide most easily. Human performance, on the other hand, contains all sorts of deviations of timing and accent dynamics.
Take a look at
/user/c/cc/220a/scm/stk/niteFunkStkMandolin-quant.scm
First it imports midi events, then separates by track and plays one on with an Stk instrument patch as before. The instrument is changed to an FM simulation of a Hammond B3 organ, hence the name "BeeThree." Have a look at the documentation for Stk's BeeThree, as well.
In the scheme code you can choose combinations of distortion of the imported events. Quantized values are rounded to a stepsize. expressed in msec. random values are first quantized then jittered randomly. Raw values are left untouched.
|
|
onset time |
duration |
amplitude |
|---|---|---|---|
|
raw |
|
|
x |
|
quantized |
x |
|
|
|
random |
|
x |
|
Turn in (by linking in the usual way) a soundfile that demonstrates quantization and randomness. This will depend on your midi source material.
* If the track or tracks that you manipulate are coming from "perfect" (score-like) values, then they are already quantized. If you are in this group, then make 3 versions as separate sections in your soundfile: 1) the original 2) a random version that sounds sort of ok and 3) a random version that is too random
* If the track or tracks that you manipulate are coming from "expressive" -- human feeling -- "real values," then make 4 sections: 1) quantized version 2) the original 3) a random version that sounds sort of ok and 4) a random version that is too random
You choose how to affect which parameters, and you can alter stepsize and random amounts by modifying the code. The sections ought to be around 30 or so seconds each. The question I'm posing with this assignment is, "can the human factor be modeled with random deviations?"