Vocoder or Convolver?

[eduardo_b]

It may simply be a lack of knowledge on my part, but don't both convolvers and vocoders use one signal to modify another? I realize they each typically have unique filtering capabilities as part of their interfaces, but do they cross over enough that one could be interchanged for the other? I'd like to understand the similarities versus differences in use from anyone more experienced with these than me before I purchase one or the other, or both. Any recommendations (and the reasons why) would be appreciated. Thanks.

[kritikon]

I suppose you can use vocoder impulses - I've never tried. The big downfall would probably be that you'd have to make your own impulses of a vocoder to get the impulses...I've not come across any (free) impulses of vocoders anywhere, although there may be commercial ones available that I don't know about. And if you make your own impulses, it means you need a vocoder to start with - so you may as well use a vocoder.

Some vocoders use modulated noise too, whereas any impulses would be static.

In terms of CPU - probably convolution would use less, as they'd be pretty short impulses on the whole. But most s/w vocoders I've used don't use that much in the way of CPU, so it's not really a big issue.

Flexibility would be an issue though. For anything vocoded, you generally want to tailor the sound quite alot to fit your desired sound, or to fit the track. Whereas with convolution you'd be stuck with the number of different impulses you have.

I'd say overall you'd have access to a wider range of sounds at present with a s/w vocoder.

[ohm]

The audio file you load in a convolver is usually pretty short (couple of seconds max). Convolution works by taking every incoming sample and process that using every single sample of the impulse (simply put). It works more like a filter than like a vocoder. In fact: it can't sound like a vocoder at all.
Convolution is commonly used to simulate the acoustic response of rooms, reverb units, microphones, speakers, eq's an other static stuff.
Voxengo pristine space is a great convolver (on windows), dunno about mac stuff yet.
edit: I just remembered that it's actually possible to perform convolution in a manner similar to a vocoder - convolution is much more than just convolving impulse responses.

Vocoders to do the "cher thing" - and they are brilliant on drums too. The process is much simpler and can not be used to simulate a room for example. No need to buy a vocoder - mda talkbox for example sounds fantastic and it's free. If you want a true analog sounding vocoder then I recommend Tassman, but it's kinda overkill just for a vocoder Btw. comb filters can be used in much the same way, except they discard the timbre of one of the signals and use just the dynamics - they are great fun though. If you have a CM mag, you have a free ohmygod comb filter on the CD.

[Music Engineer]

I'd like to understand the similarities versus differences

similarities:
both apply a frequency envelope on a signal. in convolvers this frequency envelope maybe derived by fourier-transforming the impulse response of an analog filter, for example. in most cases impulse responses are taken from acoustic environments - in this case we usually think about convolvers in the time domain. however the impulse response and the frequency envelope contain the same information - one can be derived from the other by means of the fourier-transform. convolution in the time domain is the same as multiplication in the frequency-domain (better known as "filtering") - this fact is also the basis for fast convolution algorithms.

differences:
in convolution, one signal is assumed to be an impulse response of a system which does not change over time. in vocoders in contrast, an incoming audio signal is analyzed in real time for its frequency envelope and this time-varying(!) frequency envelope is superimposed on a second signal - the classic application is to superimpose the formants of a human voice on a pad-synth like sound.
btw.: maybe you wnat to check out my free lpc-vocoder plugIn

[Music Engineer]

Vocoders to do the "cher thing"

...this was Antares Autotune

[azertopia]

As a simple user, i would say that the difference is in the accuracy.
For cross synthesis , a software vocoder have 8 to 64
bands of analysis , with the voice the result is more
"robotic"
a software convolver (you mean "morpher" perhaps ?")
with a voice the result is more "angelic", and spectrum is divided in 512,1024 or more bands
The main difference is in the technology, the morpher and convolver are in the frequency domain.
A great way to vocode in the frequency domain is to use prosoniq morph, it implicates a small latency but the sound is very nice. In the frequency domain you also have SpectrumWorks which is more flexible than morph
My recommandation is : DelayDots spectrumWorks,I keep a good recollection of the demo.

[ohm]

a software convolver (you mean "morpher" perhaps ?")

The author of "shapee" calls the process for "frequency shaping" - this is the same deal right? The sound is definately "angelic" compared to an analog vocoder.
I'm wondering what algirithm is used in mda talkbox - anybody know? Sounds pretty digital actually.
If you are running windows I'd try Talkbox and shapee before spending $$$. Spectrumworks seem like a very fun tooyl though - I too remember having fun with the demo - very flexible.

[eduardo_b]

As a simple user, i would say that the difference is in the accuracy.
For cross synthesis , a software vocoder have 8 to 64
bands of analysis , with the voice the result is more
"robotic"
a software convolver (you mean "morpher" perhaps ?")
with a voice the result is more "angelic", and spectrum is divided in 512,1024 or more bands

So I could use a convolver, substituting the IR file with a music wav file, thus achieving a softer sound than with a vocoder, correct? As an ambient guy, I'm looking for unique ways to create this kind of music, so the "angelic" result sounds more what I'd want.

The main difference is in the technology, the morpher and convolver are in the frequency domain.
A great way to vocode in the frequency domain is to use prosoniq morph, it implicates a small latency but the sound is very nice. In the frequency domain you also have SpectrumWorks which is more flexible than morph

I have the Prosoniq Morph demo and am at a loss as to how to load the two sounds. There's no indication on the plug-in for loading files and there aren't any docs with it or on their site. Any help on this would be great.
But I wasn't necessarily thinking of morhping per se. I've been reading up on vocoder history and basics and found myself thinking that the basic premise sounded like that for convolution. But I checked the archives here and didn't see any threads that discussed this, so I posted my own question. I hadn't considered the morph versus vocoder concept, so now I've got more to consider. I do have Cameleon, which does morphing.

[ohm]

So I could use a convolver, substituting the IR file with a music wav file

Not quite - there's a difference between the impulse response reverb plugins and the spectral morph plugins. The spectral morph / frequency shaping plugins I've used work like the usual vocoder plugin: one signal goes to the right channel another to the left. Perhaps the prosonic one is stereo and has two inputs? I haven't tried it.
Are you on osx or windows ?

[eduardo_b]

So I could use a convolver, substituting the IR file with a music wav file

Not quite - there's a difference between the impulse response reverb plugins and the spectral morph plugins. The spectral morph / frequency shaping plugins I've used work like the usual vocoder plugin: one signal goes to the right channel another to the left. Perhaps the prosonic one is stereo and has two inputs? I haven't tried it.
Are you on osx or windows ?

I'm a Windows person. In thinking about this, I don't think morphing is what I need. I'm looking to merge, for example, choir with pads for ambient mix, so I think vocoding would yield results much closer to what I want than would morphing. I'm in the process of acquiring a number of VSTi vocoders to try out and perhaps determine what works better. My impression is that morphing starts with a sound, blends with another sound, and then ends with the latter, whereas I'm not wanting to separate the two at any point.

[ohm]

Just make sure you check out shapee and mda talkbox - they are both vocoder-like in the way they work, and sound great

[eduardo_b]

Just make sure you check out shapee and mda talkbox - they are both vocoder-like in the way they work, and sound great

I've got them along with maybe a dozen others, some of which are way more expensive than these two. There seem to be lots of differences among them, so I'll be busy for the next few weeks. Thanks to you and others for replying.

[ohm]

It would be interesting if you posted your findings to this thread - I'm also interested in hearing how you end up using vocoders in your music.