Hello!
   I'm working on modifying voices. I wondered: Are there known recepies to 
change a male voice to a female one and vice versa/ I know I could just take 
the real thing, but that wouldn't be any fun at all for the experiment.
   I know that pitching would be involved and preferably with pvadd-family 
opcodes. Correct? But what else to do to get it right?
   Is there some statistic knowledge about vocal chords and their sounds and 
how to change specific qualities?
   Warm regards and thanks for anything
           Julien

--------
Music was my first love and it will be my last (John Miles)

======== FIND MY WEB-PROJECT AT: ========
http://ltsb.sourceforge.net
the Linux TextBased Studio guide
======= AND MY PERSONAL PAGES AT: =======
http://www.juliencoder.de

Hello,
There are really a lot of parameters that distinguish female voices from 
male ones. But for an approximation you would have to change the pitch 
and the spectral envelope independently. The spectral envelope of a 
female voice is about 20% higher than in men and the pitch is about an 
octave up. Female voices are also known to have somewhat more noise in 
the upper frequency area, so you can consider adding a small noise 
component e.g between 2500 and 6000Hz.
I have tried once to do independent modifications of pitch and spectrum 
envelope with the pv opcodes, but the results have not been very good so 
far.
You could also do this with lpc, but then you would get a kind of 70's 
computer music flavour to it à la Dodge and Lansky. It can be nice, but 
it depends what you  want.
An open program like Praat has a gender change algorithm which I think 
combines resampling (for spectral envelope shifting) and PSOLA for pitch 
shifting. It works quite well.
Best,
Andreas Bergsland
> Hello!
>   I'm working on modifying voices. I wondered: Are there known 
> recepies to change a male voice to a female one and vice versa/ I know 
> I could just take the real thing, but that wouldn't be any fun at all 
> for the experiment.
>   I know that pitching would be involved and preferably with 
> pvadd-family opcodes. Correct? But what else to do to get it right?
>   Is there some statistic knowledge about vocal chords and their 
> sounds and how to change specific qualities?
>   Warm regards and thanks for anything
>           Julien
>
> --------
> Music was my first love and it will be my last (John Miles)
>
> ======== FIND MY WEB-PROJECT AT: ========
> http://ltsb.sourceforge.net
> the Linux TextBased Studio guide
> ======= AND MY PERSONAL PAGES AT: =======
> http://www.juliencoder.de
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body 
> "unsubscribe csound"
>


-- 
Andreas Bergsland,
universitetslektor / PhD-kandidat

Musikkteknologiseksjonen
Institutt for musikk
Olavskvartalet
NTNU
7491 Trondheim

Besøksadresse: Fjordgt.1 (3.etg.)
e-post: andreas.bergsland@hf.ntnu.no
Tlf: 7359 0096
Mob: 4566 3316

Hello Andreas!
   Thanks for this information. It sounds all rather interesting. I'll try to 
havea look at praat. It's been a long time since I tried to do anything with 
it.
   Kindest regards
          Julien

--------
Music was my first love and it will be my last (John Miles)

======== FIND MY WEB-PROJECT AT: ========
http://ltsb.sourceforge.net
the Linux TextBased Studio guide
======= AND MY PERSONAL PAGES AT: =======
http://www.juliencoder.de

Hello Andreas!
   Sorry, I just replied something I meant you to read in another thread. I 
just tried praat and it's graphical, which leaves me out, as I'm blind.
   But about the spectral envelope. You must understand, I'm not too depp into 
FFT-manipulation. I do understand a bit and the rest is trial and error and 
basic deduction. :-) So how did you do it the last time you tried? How best to 
mix in the noise parts, with some cross-fading of properties (pvcross and its 
comanions) and what to do about those envelopes? I'm just getting pvshift to 
work, and I think there maybe a bit more room to optimise. Still far from 
perfect, but perhaps feasable, if all else fails.
   Kindest regards
         Julien

--------
Music was my first love and it will be my last (John Miles)

======== FIND MY WEB-PROJECT AT: ========
http://ltsb.sourceforge.net
the Linux TextBased Studio guide
======= AND MY PERSONAL PAGES AT: =======
http://www.juliencoder.de

OK so here it is. Your sample transposed up by a perfect fifth with
audiosculpt. There are three versions:

1- transposed without envelop preservation
2- transposed with envelop preservation
3- only the spectral envelop is transposed



2009/3/7 Julien Claassen :
> Hello Andreas!
>  Sorry, I just replied something I meant you to read in another thread. I
> just tried praat and it's graphical, which leaves me out, as I'm blind.
>  But about the spectral envelope. You must understand, I'm not too depp into
> FFT-manipulation. I do understand a bit and the rest is trial and error and
> basic deduction. :-) So how did you do it the last time you tried? How best
> to mix in the noise parts, with some cross-fading of properties (pvcross and
> its comanions) and what to do about those envelopes? I'm just getting
> pvshift to work, and I think there maybe a bit more room to optimise. Still
> far from perfect, but perhaps feasable, if all else fails.
>  Kindest regards
>        Julien
>
> --------
> Music was my first love and it will be my last (John Miles)
>
> ======== FIND MY WEB-PROJECT AT: ========
> http://ltsb.sourceforge.net
> the Linux TextBased Studio guide
> ======= AND MY PERSONAL PAGES AT: =======
> http://www.juliencoder.de
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> csound"
>

Thanks! It really does sound better then csound. A bit cleaner and "flatter" 
in the amplitude department.At least I guess, that it's that. :-)
   Kindest regards & thanks again
         Julien

--------
Music was my first love and it will be my last (John Miles)

======== FIND MY WEB-PROJECT AT: ========
http://ltsb.sourceforge.net
the Linux TextBased Studio guide
======= AND MY PERSONAL PAGES AT: =======
http://www.juliencoder.de

Hello Julien,
Well, actually there is a scripting interface you could use for praat as 
well which is not too complicated. I haven't used it a lot, but it would 
give you access to the PSOLA manipulation of duration and pitch which I 
thinks works quite good. And, as I wrote, there is a change gender 
function also.

As for csound, I have mainly been working with lpc where I have made an 
instrument which tries to emulate the sounds from Paul Lansky's classic 
computer piece /Six Fantasies/. Since lpc is source-filter 
analysis/resynthesis where the resynthesis is made using a synthesized 
source it is easy to add noise to the pitched source. The results are 
charming, but as for naturalness they lack a lot. I haven't really 
thought about how to add noise while using an FFT approach.

I haven't really done some serious attempts to use the pvs opcodes to 
change spectral envelope and pitch independently, but I remember that 
when trying, the results were not very good, so I am afraid that I can't 
help you too much there. If somebody has done this in a convincing way, 
I would be happy to learn what the trick is.

Best regards,
Andreas
> Hello Andreas!
>   Sorry, I just replied something I meant you to read in another 
> thread. I just tried praat and it's graphical, which leaves me out, as 
> I'm blind.
>   But about the spectral envelope. You must understand, I'm not too 
> depp into FFT-manipulation. I do understand a bit and the rest is 
> trial and error and basic deduction. :-) So how did you do it the last 
> time you tried? How best to mix in the noise parts, with some 
> cross-fading of properties (pvcross and its comanions) and what to do 
> about those envelopes? I'm just getting pvshift to work, and I think 
> there maybe a bit more room to optimise. Still far from perfect, but 
> perhaps feasable, if all else fails.
>   Kindest regards
>         Julien
>
> --------
> Music was my first love and it will be my last (John Miles)
>
> ======== FIND MY WEB-PROJECT AT: ========
> http://ltsb.sourceforge.net
> the Linux TextBased Studio guide
> ======= AND MY PERSONAL PAGES AT: =======
> http://www.juliencoder.de
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body 
> "unsubscribe csound"
>

Evening Andreas!
   I did manage someting, that takes a step in the right direction:
pitch-shift upward with factor of 2.5 and formant preservation
pitch-shift downward with a factor of something between 0.8 and 0.95 without 
formant preservation
Then use a simple channel vocoder with white-noise carrier and spoken 
pitched voice.
Filter the vocoder signal with some highpass at 2500 and lowpass around 6000.
   Mix the pitched signal and the filtered vocodeer signal.
   But it's still beyond acceptable.
   could you mail me your lpc instrument or give me a link to some sounds from 
your work. Maybe it can be helpful in my quest.
   Kindest regards
          Julien

--------
Music was my first love and it will be my last (John Miles)

======== FIND MY WEB-PROJECT AT: ========
http://ltsb.sourceforge.net
the Linux TextBased Studio guide
======= AND MY PERSONAL PAGES AT: =======
http://www.juliencoder.de

Hello again!
   About praat: Well it may be, that praat has a scripting interface, but when 
I even enter the command praat, it complains about not being able to open a 
graphic display. Perhaps the source-distro might do something for me. any 
thoughts on that?
   Kindest regards
         Julien

--------
Music was my first love and it will be my last (John Miles)

======== FIND MY WEB-PROJECT AT: ========
http://ltsb.sourceforge.net
the Linux TextBased Studio guide
======= AND MY PERSONAL PAGES AT: =======
http://www.juliencoder.de

Hello Julien,
try this link:
http://www.fon.hum.uva.nl/praat/manual/Scripting_6_9__Calling_from_the_command_line.html
Best,
Andreas
> Hello again!
>   About praat: Well it may be, that praat has a scripting interface, 
> but when I even enter the command praat, it complains about not being 
> able to open a graphic display. Perhaps the source-distro might do 
> something for me. any thoughts on that?
>   Kindest regards
>         Julien
>
> --------
> Music was my first love and it will be my last (John Miles)
>
> ======== FIND MY WEB-PROJECT AT: ========
> http://ltsb.sourceforge.net
> the Linux TextBased Studio guide
> ======= AND MY PERSONAL PAGES AT: =======
> http://www.juliencoder.de
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body 
> "unsubscribe csound"
>


-- 
Andreas Bergsland,
universitetslektor / PhD-kandidat

Musikkteknologiseksjonen
Institutt for musikk
Olavskvartalet
NTNU
7491 Trondheim

Besøksadresse: Fjordgt.1 (3.etg.)
e-post: andreas.bergsland@hf.ntnu.no
Tlf: 7359 0096
Mob: 4566 3316