more spectral mapping

Jacky wrote:

>Thanks for posting this paper. Have enjoyed it!

my pleasure...

>I've tried my hand at creating some piano samples based along
these lines... It's the "randomness" of the attack transient portion of most acoustic timbres (not to mention noise components) which seems to be quite a
beast to deal with in the realm of spectrally mapping familiar samples

yes, you need to map the partials, but also retain the "noise"
components - they add a lot of the character to a sound -
and when they are removed, everything becomes synthetic.

>What I did was to crossfade/paste mix a new sustain stage with
altered partials onto the end of the unaltered original piano
attacks.

The danger here is that the moment of transition might (or might not)
cause problems. The thing is that during the real attack, the partials
are somewhat unsteady anyway, so the mapping doesnt need to be too
exact - rather - it needs to "end right" so that the attack and the
sustain can be nicely merged. One trick I found useful was to use
an attack that was a tuiny bit longer than need be - then the
attack ends with the partials in the right place to line up with the
mapped partials in the sustain...

>What would be nice is to create at least two samples - one soft and
one bright, so that I could have a nice velocity cross-switching set-
up for the sampler.

absolutely... though this just doubles the amount of work!

> you can't stretch things too far from the original, and expect it to
sound "like" the real thing (perhaps that was my "problem" with the
above mentioned pianny).

this is part of the nature of the beast... but how far are you
moving things? a few percent should be almost imperceptible
(when heard alone), though motion of the partials by 25 percent
is almost guaranteed to introduce weirdnesses... its much like pitch
shifting algorithms - a few semitones is fone, but try to go a fifth
and you'll notice...

>But the very good news is that I can create spectrally mapped timbres
all day long which are electronic - or invented -

additive synthesis can be great

>I remember Mary mentioning a while back that she has some Mac
software that'll let one adjust the entire timbre of instruments

Metasynth is a cool program, though I still havent mastered it yet.

> Another way far out dream
is my wish for the day when computing and software power is such that
timbre and tuning *both* can be changed on the fly, and at the
composer's whim and will.

Im with you there... I'll buy one!

Bill Sethares