grain/flux balance & vocal effects

[m-laboratories.net]

OK, a non-musician friend gave me a simple task: take a vocal sample, and slow it down to the point where just a few words last 5 minutes preserving the original pitch, and WITHOUT choppy/stuttered artifacts.

I am not sure this is mathematically possible - to lengthen (loop) each phoneme the correct amount of time, and add the subtle formant shifting or other changes that would exist in a human voice saying "vvvvvvvvvvoooooooooiiiiiiiiiiccccccccceeeeeeeeeeee"

Nonetheless, I came pretty close. On my particular clip, texture setting with 12 grain and 4-6 flux seemed to work well, but it's still noticeably synthetic. Use of reverb or unison type effects to smooth the artifacting did not help.

Can you produce the desired result as though a human had actually pronounced the words that slowly? How would you do it?

[Spiralgroove]

OK, a non-musician friend gave me a simple task: take a vocal sample, and slow it down to the point where just a few words last 5 minutes preserving the original pitch, and WITHOUT choppy/stuttered artifacts.

I am not sure this is mathematically possible - to lengthen (loop) each phoneme the correct amount of time, and add the subtle formant shifting or other changes that would exist in a human voice saying "vvvvvvvvvvoooooooooiiiiiiiiiiccccccccceeeeeeeeeeee"

Nonetheless, I came pretty close. On my particular clip, texture setting with 12 grain and 4-6 flux seemed to work well, but it's still noticeably synthetic. Use of reverb or unison type effects to smooth the artifacting did not help.

Can you produce the desired result as though a human had actually pronounced the words that slowly? How would you do it?

just a though, couldnt you load the waveform into simpler and cheat a bit?
make extended pieces of each point in the wavform as loops and join them?

just a quick brainstorm...

[m-laboratories.net]

thanks for the quick response. the problem - i think - with your solution is that looping <i>any</i> small section of a vocal sample will lead to a distinctly non-human "looped" sound. I need some way of reconstructing the formant shifts & spectral changes in the looped section to make it sound like it's being produced by a human, and not the same .2 second sound being looped again and again.

Maybe a send channel with a formant presets on the EQ4 and then a unison-type detuner?

I suspect "flux" is a parameter that already does something like this, but is not specific to human voice.

[Spiralgroove]

thanks for the quick response. the problem - i think - with your solution is that looping <i>any</i> small section of a vocal sample will lead to a distinctly non-human "looped" sound. I need some way of reconstructing the formant shifts & spectral changes in the looped section to make it sound like it's being produced by a human, and not the same .2 second sound being looped again and again.

Maybe a send channel with a formant presets on the EQ4 and then a unison-type detuner?

I suspect "flux" is a parameter that already does something like this, but is not specific to human voice.

well you wouldnt wanna loop just one point... just 1 frame i guess and then move that one frame start point through the sample until you got to the last frame with a control envelope over the desired amount of time it would fill

but yeah, itd sound "loopy"

the rest of what youre talking about is a bit over my head though...

i think anything you try is gonna give you artifacting though

bummer

[m-laboratories.net]

you're probably right, there will be some artifacts no matter what. anybody else have a good idea to make this work?

[conny]

I tried both the simpler (with sample start modulated over 8 bars) and the normal timestretch method.
It seems to me that one of the objectives here is that you probably has to separate parts of the sound to get it naturally looong. I had a voice saying "Beg your leave" and it's absurd to stretch the "g" in "beg" with the same amount as the "e".
Hard task! Perhaps one would try to separate the more pure vocals from the consonants and some transitions between them and stretch these with different amounts.
There was a Swedish poet in the concretist movement who wrote a whole book by stretching the phrase "In arms" (not sure of that translation). A lot a Aaaaaa but in the same sense as above he did not put as many r's and m's etc inbetween the vocals.

// C

[LOFA]

How do you load warped audio waveforms into simpler?

[Spiralgroove]

How do you load warped audio waveforms into simpler?

drag and drop any clip into simpler

im not sure if they stay warped though

no point in having it warped if youre just taking a slice of the waveform anyway...

[conny]

The sample is not warped in Simpler but with a very short loop in Simpler and the start position moved with automation and a short midi note playing it fast this it what you get (granular...)
http://web.telia.com/~u19126355/BegYourLeave.mp3

Some reverb added.
Fade set high in simpler.

// C

[kabuki]

Here's a process I used to slow down a vocal:

Set up the vocal with the desired ammount of grain/flux. set the HiRez button ON
Pitch it DOWN one octave (-7)
Render the clip to a 24 bit wav/aif
import into iTunes - iTunes does not recognize 24 bit, and will convert it to 16... thus the octave shift.

Convert to MP3 or AIF settings high. Reconvert to aif, and back in to Live

[m-laboratories.net]

thanks everyone for thinking about this one. it drove me crazy to have bragged about how powerful the application was and not be able to fulfill his request!!

kabuki - wouldn't your suggestion change the pitch? i was hoping to find a way to slow it without artifacting and without changing the pitch.

[kabuki]

kabuki - wouldn't your suggestion change the pitch? i was hoping to find a way to slow it without artifacting and without changing the pitch.

When you convert to iTunes, it pitches it up one octave because it converts 24 bit to 16, so you pitch it down to compensate one octave before you render from Live...

[mikemc]

OK, a non-musician friend gave me a simple task: take a vocal sample, and slow it down to the point where just a few words last 5 minutes preserving the original pitch, and WITHOUT choppy/stuttered artifacts.

there is something I've messed with whereset a loop in the phonic sound of the clip if you adjust grains and rez well, zoom way in and pick a good place to loop, it doesn't glitch

So three tracks, same word, looping: beginning, middle and end: voice-

you have
vvvvvvvvvvvvvvvvvoice
voioioiiiiiiiiiiiiiiiiiiiiiiiiiiiiice
voiccccccccccccceeeee

looping on separate tracks (maybe more than three, to get beginning, middle and end Then, skootch the tracks so they are in order

vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
---------voioioiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii
-----------------------voicccccccccccccccccccce

Then draw track volume envelopes to get to here

vvvvvvv
---------oioioiiiiiiiiiiiiiii
-----------------------cccccccccccccccccccce

you could bounce it down at that point, rinse and repeat if needed and who knows you could be big in Swedish literary circles

[Vercengetorex]

When you convert to iTunes, it pitches it up one octave because it converts 24 bit to 16, so you pitch it down to compensate one octave before you render from Live...

Kabuki, a change in bit-depth will not alter pitch, only dynamic range.
It is a change in sample-rate that is affecting your exported files pitch.

[kabuki]

When you convert to iTunes, it pitches it up one octave because it converts 24 bit to 16, so you pitch it down to compensate one octave before you render from Live...

Kabuki, a change in bit-depth will not alter pitch, only dynamic range.
It is a change in sample-rate that is affecting your exported files pitch.

I stand corrected.

Do THAT!

(It all happened by accedent. All I know is it sounded clean...