Re: OT: True Sound Morphing

[Mech <mech@m3ch.net>]

At 5:21 PM +0100 2/15/09, Per Boysen wrote:
>On Sun, Feb 15, 2009 at 4:54 PM, mark francombe 
><mark@markfrancombe.com> wrote:
>>  So musically, would it be possible to morph one frequency into 
>another? For
>>  example, if we divided the frequency spectrum into... 5 sections 
>(Bass, Low
>>  mid; Mid; Upper Mid, Treble) analogeous with the "points" visually... 
>and
>>  within these frequency bands analalysed, frequency pitch.... 
>NOPE...Thinking
>>  aloud here... Im not thinking this would work...
>
>
>Yup,  I tend to agree that musical morphing is not possible. Not
>physically in sound, that is.

Well, two quick points that haven't yet been addressed (I don't believe).

First, actual physical morphing of the type Mark mentions should be 
entirely possible using resynthesis (breaking an analyzed sound down 
into its base sine waves, then manipulating those 
frequencies/durations in exactly the way one "morphs" parameters on 
the Vortex).

As always, this runs into two immediate issues: quality and 
feasibility.  The former comes into play in that most sounds (much 
less musical passages) are just silly complex and hard to bust down 
to additive synthesis while still retaining their original 
personality.  Most attempts sound more like a bad vocoding job.  To 
the latter point, resynthesis is, at this point in time, still a 
bloody PITA.  I can remember programs like Lyre (on the PC) or Avalon 
(for Atari) back in the 80's, where you would feed in a one-second 
sample then go to bed for the evening 'cause it would take several 
hours just to chew on it.  Today's computers are much more powerful, 
but the process is still the same: wait for the computer to analyze 
the sample, then see if the finished product is usable at all.

I've got the Hartmann Neuron VST (along with the Wavemaker software) 
and it does about as good a job as any I've seen at producing 
resynthesized models that you can morph between.  However, you 
quickly become sidetracked, as it becomes apparent that its far more 
interesting to manipulate the models in all sorts of unforeseen ways 
than to simply morph between two resynthesized sounds.

Second, perhaps a more realistic (not to mention practical) method of 
morphing is to instead use a Phase Vocode or Spectral Impression 
algorithm to overlay the spectral qualities of one signal onto 
another.  IIRC, most of these tools (Tom Erbe's Soundhack, for 
instance) also do their processing offline, so they aren't the best 
for live sources.

However, one program that's been around for two decades (in one form 
or another: first for the Atari, then Amiga, SGI, PC, and now finally 
Intel Mac) is Prosoniq's Morph: 
http://www.prosoniq.com/main/prosoniq-morph/

I used to use it years ago on the PC, and with the recent release of 
the AU version it's been floating around the top of the "I need to 
get this when I have some cash" list for most of a year now.  Does a 
very good job of approximating a good audio morph.  Although it's not 
perfect (nothing is), IMNSHO it's worth downloading the demo to see 
if it's anywhere close to what you're looking for, Per.  :)

        --m.
-- 
_____
"bye-bye empire; empire, bye-bye"