posted 01 March 2002 12:57            

This will impose the spectral envelope of the speech onto the wind and will sound like the wind is talking.

In general, the spectral analysis/resynthesis is not well-suited to the resynthesis of broadband noisy signals like wind or water or noise. It is trying to model the signal as the sum of sine waves, so unless you have an infinite number of sine waves to work with, you can end up with gaps in the spectrum (and a signal that sounds as if it has been comb-filtered).

posted 03 March 2002 14:40            

try the following: take an empty glas and hold it between your hands in front of your mouth. fold your hands like you want to wash your face with some water. the result will appear like a gas mask.
experiment with it; speak/whisper into it like you imagine your speaking/whispering wind (and donīt forget to record it, of course;-). the result already will be quite effective on its own.
use that as the source for your vocoding/re-synthesis. because it already contains the gesture of the speaking wind, the final result will sound much more "natural".
i found out that the creativity and perfomance *before* using kyma is the key to convincing sounddesign. kyma is the final executor for our ideas, not the creator. (thatīs speaking for every kind of equipment!)
again technically: the more similar the source signals are, the more convincing will any morphing/crossfading/blending/warping/miracling be.