Re: [Octal-dev] Fourier Transforms and samples

[Aaron Thieme]

> closely, so I can see this for myself. Anyhow, what I'm getting at is
> this: would it be worth it to store selected samples as transformed? Less

Funny, I've been thinking about this  (at a conceptual, not code) level
for the last few weeks, for a related project.

However I was thinking of producing something more like a sampler 
instrument building helper, than a real time cache.  

Question: how to produce a minimally useable "instrument" from a sample or
two, and minimize the Casio SK1 affect -- which I'm guessing is "formant"
related, not really knowing what that is.

I invisioned allowing a user to specify a number of "seed" samples which
could be mapped to pitches.  The user would then specify the scale to be
generated (12 tone, pentatonic, one of the gamelan scales...).  The most
useful range would obviously be between the available seeds...

Intermediate pitches would be generated by pitchshifting alone with a
certain distance (say 200%) from a "seed".  The trick would be to apply 
convolution between *two* pitch shifted seeds to fabricate samples for
pitches > 200% from either.  I imagined that the degree of convolution,
mix varied by distance to seeds (linear..?), etc., could be dialed in by
the user...

My thinking here was that this might crudely produce a more natural
intermediate value than pure shifting on a given seed.

Also, that it it would be cool to quickly produce non-conventional tuned
"instruments."

Also, use of unusual "seeds" (completely different noise samples, bits of
ambient sound, etc.) in the set might produce a cool set of
blended/convolved textures that, when played across a conventional pitch
stream, might be high in serendipitous interest...

A bit off topic, but I rarely have anything to contribute... :)

aaron

  address@hidden
  http://www.quietamerican.org