Time/Frequency Representations from an Artistic Point of View
| Date | 1999-03-04 00:47 |
| From | Michael Gogins |
| Subject | Time/Frequency Representations from an Artistic Point of View |
|
One of the cool things you can do in computer
music is time/frequency morphing. I am interested in it for several purposes,
but the usual techniques don't satisfy me. I'd appreciate some feedback or
discussion of the underlying issues.
What I want to do, without introducing
artifacts, is:
Stretch or shrink time without affecting
pitch.
Transpose pitch without affecting
time.
Combine stretching and shrinking of time with
transposition of pitch.
Translate visual images into
sounds.
Translate sounds into particularly intelligible
visual images.
Generate fractals and
translate them into sounds.
Color one sound with another, perhaps with the
two sounds running on different scales of time or pitch.
I have done all of the above, sometimes in
several different ways, usually with Cool Edit Pro or Csound, sometimes with
software written by me. So, I think I have learned some
things:
The phase vocoder can be
exact for some transformations, but introduces artifacts for many
transformations.
Filterbank
analysis/resynthesis is not exact, but does not introduce artifacts if enough
filters and oscillators are used, and can in fact be extremely realistic, more
so in practice than the phase vocoder. However, it does not represent sound
using any sort of regular grid into which a fractal or image could be quantized
in an intuitive way.
It seems that sonograms are effective pictures of sounds, but
making sonograms and translating them back into sound is not really a very good
way of working.
The acoustical and musical imagination morphs mental images of
sounds in some extremely sophisticated way that keeps phase effects, smearing
from convolution, and so on completely in the background. It seems to be
possible for the mind to switch back and forth between a "grid" or
"sonogram" picture of sound and a "harmonic" or
"filterbank" picture of sound without even thinking about
it.
To map a fractal or image onto a time/frequency/phase grid almost
invariably produces a startling array of artifacts including clicks, ringing
sounds, smearing, and burbling. However, when I look at a picture of the fractal
or at the image, I can easily imagine what I want the damned thing to sound
like.
In fact I, can ALREADY do this by translating
hetro analysis files into scores of notes that I synthesize using slurs, so that
the result for each track of the filterbank analysis is a connected track of
sinusoidal sound. However, this technique does NOT lend itself to representation
of fractals. Nor does it allow one to color one sound with another except by (a)
mixing or (b) convolving the oscillator signal with some impulse.
I also want to color sounds with each other in ways that do
not produce convolution artifacts. For example, I would like to record the sound
of the wind blowing through the leaves of trees on a summer evening and color it
with some pitched sounds or with voices, or give the sounds themselves some sort
of pitch, perhaps by bringing out pitches that are inherent in the sounds
without otherwise distorting their timbre. I can IMAGINE the results but I do
NOT know how to compute them.
Some of the above is probably confused or uninformed, but -
any help?
|
| Date | 1999-03-04 05:49 |
| From | Ricardo MadGello |
| Subject | RE: Time/Frequency Representations from an Artistic Point of View |
|
Sometimes a vacation back in old analog-land with an ear to what's going
on around one's self can break these chains.
A
train yard or airport has plenty of time/frequency morphing to feed
this.
The
thrill of a siren screaming by can be enough to unclog the ear
canals.
A
flock of geese taking off from the lake in front of you leading into the slap of
their wing tips as they fly overhead calling their joy of flight to the world as
they breeze by is even better.
And
yes, this is csound related as this last bit is central to a theme I'm slowly
gathering wool for production of in a performance space. The only tool I
know that can allow me to achieve this, aside from lots of money for tons of
amps and speakers,etc. ;-), is csound.
Back
to basics.
Everything i know is wrong when the ideas aren't
poppin'.
Ricardo MadGello
Out & About.. .
. . .
. .
.
-----Original Message-----
From: owner-csound-outgoing@maths.ex.ac.uk [mailto:owner-csound-outgoing@maths.ex.ac.uk]On Behalf Of Michael Gogins Sent: Wednesday, March 03, 1999 4:47 PM To: csound@maths.ex.ac.uk; music-dsp@shoko.calarts.edu Subject: Time/Frequency Representations from an Artistic Point of View One of the cool things you can do in computer
music is time/frequency morphing
:
: : Some of the above is probably confused or uninformed, but -
any help?
|