Yes I have actually used a seamless loop of a person who can growl quite well as the modulator and as you may expect it sounds precisely like the noise signal modulation with the addition of a bandpass (due to the human formant).
Growling is by nature a chaotic signal with high dynamic range, which is precisely what noise is.
The information I seek is perhaps similar to a lot of other FFT vocoders that take many steps in producing their intended result.
I am looking to expand my knowledge of using FFT to vocode and looking beyond just the simple multiplication of 2 samples amplitudes to produce an effect.
I need more information on how one signal can be used to modify another in either the frequency domain or amplitude domain (or ways to combine them differently)