Vocoder as "autotuner"

[Krakatau]

You may be true, but perhaps it would make sense if you can pick a waveform that is as close as an "unfiltered sound by the thoat and mouth's articulation" should be...a few, uncolored , vocal string's samples ("cordes vocales" in frensh)

( Don't think about it too long though... )

[Music Engineer]

You may be true, but perhaps it would make sense if you can pick a waveform that is as close as an "unfiltered sound by the thoat and mouth's articulation" should be...a few, uncolored , vocal string's samples ("cordes vocales" in frensh)

maybe a sound of unfiltered vocal cords can be created by inverse filtering the some voice. inverse filtering is an operation which involves adaptive filtering in a way my LPC-vocoder freebie does. it's also called whitening. you may try it by feeding a voice (as carrier in this case) into the LPC-vocoder, turn up the "whCar" (whitened carrier) volume in the mixer (set "Voco" to zero). this should give you a whitened version of the carrier input. maybe some tweaking of the whitening filters parameters makes sense

[meister eder]

(- Is it possible then to pick some waveform up from inside this voice take and choose them as carrier waves ?
- would it make sense if the vocoder allow it ? )

...can VOCATOR do something like this ?

of course you could pick the wavefrom in a sample editor, load it into a sampler (as looped waveform) and then use the sampler output as carrier. but i don't think that this would make much sense - the waveform of the voice already contains the vocal formants of the particular vocal where the waveform is taken from. whereas in vocoding we usually want some "neutral" broadband signal as carrier (neutral in the sense, that it doesn't have it's own strong formants)

Taking spectrum from the voice:
Like braindoc says, it's possible in a sample editor. Not all sample editors provide the ability to extract an average spectrum though. If you are on Mac, then Metasynth does this perfectly. You can save the spactrum as an audio file.
Another trick is to use the "Hold" function of Smart Electronix Ambience. Just hit it in the right moment, make sure it's recorded as automation data, and then render the result. With that method you get a quite broadband spectrum with some tonal information. What you capture is just one moment of the original audio though.

"Can Vokator do something like this?":
The Carrier section of Vocato can be a synth which can use a sample instead of oscillators, too. So it should be possible easily. You can then also experiment with the inbuilt spectral effects, e.g. with something like "smear", an effect that - like the name says - blurs the spectrum of the carrier signal.

Neutral broadband signal as carrier:
Fact is, that the output signal of a Vocoder consists only of frequencies that are inherent in both the carrier and the modulator signal. So logically it would indeed make sense to take an average spectrum of the original voice as the carrier. Overpronounciation of some formants might be a problem though, right. But you could correct that with the inbuilt equalizer of the Vocoder to some extend.
The thing is, if you take a signal that is too broadband (e.g. noise), then you won't be able to play it melodically. If you take a signal that is too alien in means of spectrum, then the sonic result will be artificial.
Ergo: I think for a "transparent" sound it could be a good idea to experiment with a spectrum from the voice as carrier.

[Branis]

Maybe you don't need a vocoder, but something like Choir Boy from Elevayta. You can use it's MIDI mode to control the pitch of your vocal, and it can also preserve formants which is very important if you want a realistic sound.

[meister eder]

maybe a sound of unfiltered vocal cords can be created by inverse filtering the some voice. inverse filtering is an operation which involves adaptive filtering in a way my LPC-vocoder freebie does. it's also called whitening. you may try it by feeding a voice (as carrier in this case) into the LPC-vocoder, turn up the "whCar" (whitened carrier) volume in the mixer (set "Voco" to zero). this should give you a whitened version of the carrier input. maybe some tweaking of the whitening filters parameters makes sense

"Inverse filtering" and "Adaptive filtering". Never heard of that before, but sounds like interesting concepts.
Can you explain a bit further what this does? I didn't quite get it.
Does it something like generating a white noise signal out of the source?
And what exactly gets inversed in the inverse filtering process?

A hint to some technical explanation on the web would be much appreciated!

[Pouckypluysch]

Also, you can autotune things manually in things like Cubase or Pro Tools.

U sure? I haven't seen that function. Is it done with the pitch shifter?

[Music Engineer]

"Inverse filtering" and "Adaptive filtering". Never heard of that before, but sounds like interesting concepts.
Can you explain a bit further what this does? I didn't quite get it.

meister eder - sounds germish. if you understand german and you are a little bit into math and digital signal processing, i can send you a paper, i have written in the context of some lecture, i had at university (it's written in german).

Does it something like generating a white noise signal out of the source?
And what exactly gets inversed in the inverse filtering process?

depends on the input signal. if it is coloured noise, then - yes - the output will be white noise. if the input is some signal with a harmonic spectrum, then the output will be some kind of impulse train. it always tries to make the output signal as white as possible - but it of course can't create any frequencies which aren't present in the input signal (it's filtering after all). what is getting inversed is a filter in a - so called - source-filter-model of the signal. read the manual of my LPC-vocoder plugIn for more info (although i don't go too much into the details, there).

A hint to some technical explanation on the web would be much appreciated!

eerrrm, OK. i will look for my paper on my HD and put it online. you can also google for "linear prediction", "LPC", "adaptive filtering" or something like that

[Music Engineer]

A hint to some technical explanation on the web would be much appreciated!

here is my paper:

www.braindoc.de/documents/Formantanalyse.pdf

[meister eder]

Thanks, braindoc, interesting read.
You really are a braindoc!

[markanini]

Well I looked at all the vocoders I could find in the KVR database and none fullfilled my needs. And vokator sure was overwhelming. Does anyone that uses vokator here have any ideas on what settings to use or have a preset?

[Krakatau]

Well I looked at all the vocoders I could find in the KVR database and none fullfilled my needs. And vokator sure was overwhelming. Does anyone that uses vokator here have any ideas on what settings to use or have a preset?

You'd need to create your own patch in wish you'd use a single waveform of your original voicetake processed in a way that all the filtering aspects created by the throat and mouth would be removed ( "inverse filtering" as said before in the thread)

Something that vokator itself should not be able to do

...but to be honest, this is leaving me pretty because i canno't figure out how any software could analyse a voicetake's signal deeper enough to make clear distinction between vocal strings (right tranlation ?) sound generation( with it entire harmonic content in itself ) and a model of other part of the body involved into the singing process

All what i really understand ( do i really ?) about what i've read show a kind of process that after analyse of a single waveform, would rise/decrease all the resulting sinewaves that composes the fondamental and harmonic content of the signal ( the way a coloured noise would "turn into white" )

_____

( P.S. : Forgive me because, i left my study 23 ears ago by now and my english is sometimes restricted )

[sonicfire]

MDA talkbox feeded with plain saw and some white noise!

[> DiGiT <]

try klang labs vokko... maybe?