I READ A VERY INTERESTING INTERVIEW WITH JIM O'ROURKE YESTERDAY IN THE MUSIC NEWSLETTER TONE GLOW. THERE WAS QUITE A LOT OF MATERIAL TO LEAN ON SO I SHAN'T USE IT ALL NOW. PERHAPS IT WILL REFRAME SOME THINGS IN THE FUTURE, AND HAS CLARIFIED A FEW SUSPICIONS OF THE PAST.
HAD I HEARD STEAMROOM 47 PRIOR TO READING IT I'D HAVE BEEN POSITIVE ABOUT IT ALL THE SAME: IT'S PROBABLY MY FAVOURITE STEAMROOM SO FAR THAT WASN'T OUT AS SOMETHING ELSE ALREADY. IT PRESENTS FRACTURED PIANO PLAYING INTERMESHING WITH ELECTRONICS. THE PLAYING ITSELF IS STRANGE AND ODDLY BRITTLE - LIKE A FUMBLING ATTEMPT TO DEVELOP A BAROQUE MELODY THAT COLLAPSES (BUT INTERESTINGLY) AND RISES AND FALLS. IT IS INTERESTING STUFF.
THE INTERVIEW PROVIDES SOME INSIGHT ON HOW THIS PARTICULAR RELEASE WAS MADE. REMEMBER WHEN I SUGGESTED A LONG TIME AGO THAT PROCESS AND HOW THE SOUND WAS MADE FEELS OF PRIMARY IMPORTANCE RATHER THAN THE END PRODUCT OR HOW IT MIGHT MAKE YOU FEEL? WELL:
Well there's this whole area of research, this grouping of disciplines called MIR—music information retrieval.
I'm not familiar with it.
The results of that are those things on your phone where you can ask, “What’s that song?” and figure out what it is.
Oh, apps like SoundHound?
Yeah, so basically that world of research is called MIR. Which is a combination of other disciplines like FFT (Fast Fourier transform) analysis and delta signals and all that shit. As a sort of adjunct to my near-lifelong obsession with Roland Kayn
and the idea of music based on cybernetics, I was researching various
things about MIR over the past couple years—not because I'm interested
in it as a technology, but because of the parts of technology that make
it up.
What that album is, and the name’s sort of a clue—I’m
surprised no one’s caught on, it’s not that obscure. But basically I
made an FFT analysis of—
Glenn Gould?
Yeah, Glenn Gould! So I did an FFT analysis of all the Goldberg Variations.
But what I did was, for each particular variation, based on the simple
ideas of—and they were simple because I was just doing tests—how many
modulations and blah blah blah. When you're doing FFT analysis, it's basically like a time domain
analysis. Depending on the sample rate you're working at, you’re
cutting up music into slices of time, almost like plastic overlays. It’s
kind of like when you see those things when you make slices of the
brain and you’re putting see-through, plastic pieces on top of each
other for each slice.
So with the Fourier transform, you're taking
time slices and you're reading the energy levels of the frequencies as
sine waves—sinusoidal—for each slice. Then to recreate it, these slices
and the sine waves and particular amplitudes are recreated in the same
time domain. But what I was doing was changing the time domain as well
as resynthesizing it—not with sine waves, but with a wave table of a
hundred and fifty different wave forms that were dependent upon both the
frequency and the amplitude in regards to something particular about that Goldberg variation.
So I was recreating those performances of the Goldberg Variations with this kind of skewered, nutso, FFT resynthesis. And then what I did (laughs) was once I resynthesized those—which is that kind of crazy, squelchy, noisy part—I ran that through a program which tries to read musical material out of the audio. And that made a new score that was played on the
SO THE PIECE ENDS UP BEING THIS TWICE REMOVED GOLDBERG VARIATION THAT PERHAPS TRUMPS THE IDEA OF ALGORITHMS WRITING OUR FUTURE MASTERPIECES BY DOING THE THING HUMANS DO BEST AND TWISTING IT AWAY FROM OBVIOUS LEVELS OF INTENT.
THE THING IS THOUGH IT SOUNDS REALLY GOOD. AS STATED - I AM NOT A PROCESS GUY. I LIKE EXPRESSION AND RESULT. FORTUNATELY JIM'S GOALS AND MINE ALIGN HERE FOR AN EASY CERTIFIED: GOOD!