Rough Raga Riffs
hardware:
NeXT machine
software:
cmix,
rt
There is a story I tell about the genesis of this piece:
During my commute up to New York from my home in Roosevelt, I often
check out several different radio stations. Two in particular were
very close to each other on the FM dial: WFDU from Farleigh-Dickinson
University (the "folky" one) and WFMU, an independent station in
East Orange, NJ ("pirate" radio!). At the time of my commute,
WFDU featured a 'music from around the world' show, and WFMU
("pirate" radio!) was playing it's typical speed-metal
("pirate" radio!) mish-mosh. I have no real musical taste,
so I enjoyed listening to both of them.
At the time this was happening, I was also working to figure out
how best to use Charlie Sullivan's modified Karplus-Strong synthesis
instrument. It had a number of interacting non-linear parameters that
made it difficult to control. I was heeding Paul Lansky's advice
about testing synthesis algorithms and imbedding the tests in a
"musical context", although I had been simply trying to produce
rather vanilla scales with the algorithm.
One day I was listening to some amazing sitar piece on WFDU, and
flipping between the blindingly-fast sitar "riffs" and a Megadeth
piece on WFMU ("pirate" radio!) with blindingly-fast guitar
"riffs". Hey! I decided to try to merge the two in my STRUM
(the name of the cmix virtual instrument based on Charlie's code)
tests -- a heavy-metal sitar player, or a South Indian speed-metal
guitarist. Rough Raga Riffs is the result.
You can actually hear the evolution of the style-modeling rules
I encoded through the piece. For example, about a third of the
way through I 'taught' the style-model how to do the Eddy Van Halen
trick of tapping on the guitar strings to produce a fast,
repeating sequence of notes. I find this co-development of code
and music to be a very congenial way to create music.
See the
commentary
about my piece Almost Real for more observations about
style-modeling. There are a couple of things I would probably
change about Rough Raga Riffs if I did it over: one is
to slightly shorten the roller-coaster-like riffs towards the end
of the piece, and the other (which I should probably correct
anyhow) would be to remove a strong DC bias component that compromises
the dynamic range of the piece and makes it a lot softer than
other recordings.
When I did this piece it typically took hours to render a musical
passage. The basic code is now included as a demo of my max/msp
external object
[maxlisp]
object, running in real-time without causing the processor to break
a sweat.
The original LISP source code I used to create Rough Raga Riffs
is available
here.