(hang in there Victor - your cue comes up in a bit..)

Following on from all this...

http://www.nabble.com/Pitch-Syncronous-Granular-Synthesis-%28or-does-Curtis-Roads-dream-of-electric-Britneys%29-tt14707962.html


This is one of the more "authentic sounding" implementations of a "formant
tweeky" vocal effect i have heard. 

http://www.xponaut.com/showpage?pid=91

But it still lacks the basic ability to "wilfully impose pitch" rather than
just being a correction tool.

The manual doesn't go into much detail, but gives some vague insights, &
alludes to FFT being in the mix somewhere (for tracking & formant
identification assumedly) as well as use of the key words "pitch syncronous"
in an otherwise evasive & undetailed context... mmm, how does all this fit
together? (i'm as determined as hell to see this realised - this effect has
eluded me for years!)

The GUI also looks a bit like what i'm sure the original "suitcase synth"
that did this looked like... (does anybody know what that synth is?)

The other thing i thought in response to all this (Victor this is your bit)
is that if pvscale can shift pitch & preserve formants - why couldn't the
same algorithm be used to create access to "formant tweaking"?? - again in
combination with pvspitch tracking & correction, this might provide
something pretty useable & controllable..)

So Victor, do you have any thoughts on this? does "pvsformant" sound like a
reasonable opcode request? What do you know of more "conventional" pitch &
formant shifting effects? am i missing something obvious already in the
csound arsenal? (Basically i want to repitch & formant tweak vocals without
"Vocoding" them...)

I got partikkel functioning basically this arvo (in 5.06) & playing back a
vocal sample... but it's still a long way short of doing all the formant
identification & "effective resynthesis" I'm after - (not implying it can't
be done - but its still a long road... as my original quotations from
Computer Music tutorial suggest - I still don't know what the hell this
means

"The output signal results from the excitation of the pulse train on the
weighted sum of the impulse responses of all the filters" (conclusion of
Curtis Roads description of Pitch Syncronous Granular in Computer Music
Tutorial page 174-175)

Does anyone know of the articles i refered to that are referenced in
Computer Music Tutorial?

They include an MIT press publication 1991 - "Representations of Musical
Signals..."

If someone can assure me that the process is described in sufficient detail
in Microsound that it could be implemented based theron i'm prepeared to
splurge & grab it from Amazon... but really i'm still looking for one of the
"king hitter developers" to help me over the line with this if at all
possible...

& when the hordes of spotty, DSP obsessed, WASP males come flooding in via
csounds.com you can thank me later...


-- 
View this message in context: http://www.nabble.com/Britney-Spears-of-Destiny-%28esp.-attn-Victor-Lazzarini-%2B-others%29-tp14730945p14730945.html
Sent from the Csound - General mailing list archive at Nabble.com.

>
>The other thing i thought in response to all this (Victor this is your bit)
>is that if pvscale can shift pitch & preserve formants - why couldn't the
>same algorithm be used to create access to "formant tweaking"?? - again in
>combination with pvspitch tracking & correction, this might provide
>something pretty useable & controllable..)
>
>So Victor, do you have any thoughts on this? does "pvsformant" sound like a
>reasonable opcode request? What do you know of more "conventional" pitch &
>formant shifting effects? am i missing something obvious already in the
>csound arsenal? (Basically i want to repitch & formant tweak vocals without
>"Vocoding" them...)

Well, formant tracking and preservation is a complex business.  What pvscale
does is quite simple (and fast): it does not shift the magnitudes along 
with the
frequencies, which works for some sounds, but not for others.  Component
frequencies might be moved to dips in the spectrum, where they disappear,
so it sometimes doesn't work as well. But as I showed 
(http://www.dafx.ca/proceedings/papers/p_275.pdf),
it can work well, and very cheaply. The guys at IRCAM have developed a good
formant preservation algorithm, but it's expensive and they are very snotty
about cheap tricks such as the one I did. If you want to read their paper, it's
online at: http://dafx05.ssr.upm.es/Proc_DAFx05/P_030.pdf

Victor


Victor Lazzarini
Music Technology Laboratory
Music Department
National University of Ireland, Maynooth

>If someone can assure me that the process is described in sufficient detail
>in Microsound that it could be implemented based theron i'm prepeared to
>splurge & grab it from Amazon...

What happened to the library? ;) I can tell you as soon as I get home.

Alex

Sunrise ADSL mit automatischer Installation - die einfachste Art, ins schnelle
Internet einzusteigen. Ab CHF 9.90 pro Monat. http://www.sunrise.ch/adsl

Microsound is a terrific book that I would highly recommend; however,  
while it will provide you with all of the theory you would need to  
create a Britney Spears of your own, it won't walk you through the  
actual implementation.

But you should buy it (or borrow it) nonetheless-- it's been an  
invaluable addition to my personal library.  The chart on time-scales  
in music alone is worth the price of admission.

JP


On Jan 10, 2008, at 7:52 AM, alexweiss@freesurf.ch wrote:

>> If someone can assure me that the process is described in  
>> sufficient detail
>> in Microsound that it could be implemented based theron i'm  
>> prepeared to
>> splurge & grab it from Amazon...
>
> What happened to the library? ;) I can tell you as soon as I get home.
>
> Alex
>
> Sunrise ADSL mit automatischer Installation - die einfachste Art,  
> ins schnelle
> Internet einzusteigen. Ab CHF 9.90 pro Monat. http://www.sunrise.ch/ 
> adsl
>
>
>
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body  
> "unsubscribe csound"

Your paper looks great. i might even understand a lot of it! 

& now I see you also wrote syncgrain... How far away is that from being
implemented as "pitch syncronous" (potentially), as opposed to just
"syncronous"

You surely must be on top of the Pitch Syncronous Granular Synthesis process
as Roads describes it in Computer Music tutorial? 

Everything i've encountered about this seems to sooner or later relate
"Pitch Syncronous" & "Formants" together as happy bedfellows in all this....
& that is definately what i'm trying to get at. & you've already got your
finger in both pies victor!

Even my Loris ambitions & experiments with SDIF .txt files all come back to
the same recurring theme....

1) Spectral analysis (FFT or SDIF output - could be loris)

2) "meaningful abstraction" of that data into a) fundamental, b) "band
enhanced overtones" / "formants", & c) noise

3) manipulations & interpolations based on 2 above 

that's possibly an aside, but with so much firepower (potentially) at my
disposal via csound it does truly seem like the (next) missing link.... even
ADSYN & LPANAL / LPRESON offer the same "model" but not the flexibility to
"use" the data in this synonymous way... (effectively im talking about
implementing the same technique here 5 or 6 ways... does that make any
sense, is there any value in entertaining these ideas?) 

But let's keep it simple - do you know how i could attempt PSGS in Csound? 


Victor.Lazzarini wrote:
> 
> 
>>
>>The other thing i thought in response to all this (Victor this is your
bit)
>>is that if pvscale can shift pitch & preserve formants - why couldn't the
>>same algorithm be used to create access to "formant tweaking"?? - again in
>>combination with pvspitch tracking & correction, this might provide
>>something pretty useable & controllable..)
>>
>>So Victor, do you have any thoughts on this? does "pvsformant" sound like
a
>>reasonable opcode request? What do you know of more "conventional" pitch &
>>formant shifting effects? am i missing something obvious already in the
>>csound arsenal? (Basically i want to repitch & formant tweak vocals
without
>>"Vocoding" them...)
> 
> Well, formant tracking and preservation is a complex business.  What
> pvscale
> does is quite simple (and fast): it does not shift the magnitudes along 
> with the
> frequencies, which works for some sounds, but not for others.  Component
> frequencies might be moved to dips in the spectrum, where they disappear,
> so it sometimes doesn't work as well. But as I showed 
> (http://www.dafx.ca/proceedings/papers/p_275.pdf),
> it can work well, and very cheaply. The guys at IRCAM have developed a
> good
> formant preservation algorithm, but it's expensive and they are very
> snotty
> about cheap tricks such as the one I did. If you want to read their paper,
> it's
> online at: http://dafx05.ssr.upm.es/Proc_DAFx05/P_030.pdf
> 
> Victor
> 
> 
> Victor Lazzarini
> Music Technology Laboratory
> Music Department
> National University of Ireland, Maynooth
> 
> 
> 
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> csound"
> 
> 

-- 
View this message in context: http://www.nabble.com/Britney-Spears-of-Destiny-%28esp.-attn-Victor-Lazzarini-%2B-others%29-tp14730945p14733136.html
Sent from the Csound - General mailing list archive at Nabble.com.

Just some thoughts around the questions of formant tracking from the 
earlier PSGS / Britney discussion:
Would it be possible to base formant tracking algorithm upon lpc? Speech 
analysis software like Praat have formant tracking based on lpc 
including formant manipulation.
At least LPC can produce nice and smooth spectral envelopes, which I 
don't think it is too complex to do peak (and maybe also valley?) 
picking from.
This could produce an output of formant (and antiformant) frequency 
values, bandwidths and amplitudes, and maybe this could somehow be used 
to control the amplitudes of the pvs?

Andreas Bergsland

>> The other thing i thought in response to all this (Victor this is 
>> your bit)
>> is that if pvscale can shift pitch & preserve formants - why couldn't 
>> the
>> same algorithm be used to create access to "formant tweaking"?? - 
>> again in
>> combination with pvspitch tracking & correction, this might provide
>> something pretty useable & controllable..)
>>
>> So Victor, do you have any thoughts on this? does "pvsformant" sound 
>> like a
>> reasonable opcode request? What do you know of more "conventional" 
>> pitch &
>> formant shifting effects? am i missing something obvious already in the
>> csound arsenal? (Basically i want to repitch & formant tweak vocals 
>> without
>> "Vocoding" them...)
>
> Well, formant tracking and preservation is a complex business.  What 
> pvscale
> does is quite simple (and fast): it does not shift the magnitudes 
> along with the
> frequencies, which works for some sounds, but not for others.  Component
> frequencies might be moved to dips in the spectrum, where they disappear,
> so it sometimes doesn't work as well. But as I showed 
> (http://www.dafx.ca/proceedings/papers/p_275.pdf),
> it can work well, and very cheaply. The guys at IRCAM have developed a 
> good
> formant preservation algorithm, but it's expensive and they are very 
> snotty
> about cheap tricks such as the one I did. If you want to read their 
> paper, it's
> online at: http://dafx05.ssr.upm.es/Proc_DAFx05/P_030.pdf
>
> Victor
>
>
> Victor Lazzarini
> Music Technology Laboratory
> Music Department
> National University of Ireland, Maynooth
>
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body 
> "unsubscribe csound"
>


-- 
Andreas Bergsland,
stipendiat 
Institutt for Musikk
HF-fakultetet
NTNU
7491 Olavshallen

E-post: andreas.bergsland@hf.ntnu.no
Tel : 73 59 73 55 
Mob : 45 66 33 16 

As far as I understand Roads' terminology, fof, fof2, fog and
syncgrain/diskgrain/syncloop would allow for pitch-synchronous GS.
So I guess these would be your starting point.

Victor

At 13:41 10/01/2008, you wrote:

>Your paper looks great. i might even understand a lot of it!
>
>& now I see you also wrote syncgrain... How far away is that from being
>implemented as "pitch syncronous" (potentially), as opposed to just
>"syncronous"
>
>You surely must be on top of the Pitch Syncronous Granular Synthesis process
>as Roads describes it in Computer Music tutorial?
>
>Everything i've encountered about this seems to sooner or later relate
>"Pitch Syncronous" & "Formants" together as happy bedfellows in all this....
>& that is definately what i'm trying to get at. & you've already got your
>finger in both pies victor!
>
>Even my Loris ambitions & experiments with SDIF .txt files all come back to
>the same recurring theme....
>
>1) Spectral analysis (FFT or SDIF output - could be loris)
>
>2) "meaningful abstraction" of that data into a) fundamental, b) "band
>enhanced overtones" / "formants", & c) noise
>
>3) manipulations & interpolations based on 2 above
>
>that's possibly an aside, but with so much firepower (potentially) at my
>disposal via csound it does truly seem like the (next) missing link.... even
>ADSYN & LPANAL / LPRESON offer the same "model" but not the flexibility to
>"use" the data in this synonymous way... (effectively im talking about
>implementing the same technique here 5 or 6 ways... does that make any
>sense, is there any value in entertaining these ideas?)
>
>But let's keep it simple - do you know how i could attempt PSGS in Csound?
>
>
>Victor.Lazzarini wrote:
> >
> >
> >>
> >>The other thing i thought in response to all this (Victor this is your
>bit)
> >>is that if pvscale can shift pitch & preserve formants - why couldn't the
> >>same algorithm be used to create access to "formant tweaking"?? - again in
> >>combination with pvspitch tracking & correction, this might provide
> >>something pretty useable & controllable..)
> >>
> >>So Victor, do you have any thoughts on this? does "pvsformant" sound like
>a
> >>reasonable opcode request? What do you know of more "conventional" pitch &
> >>formant shifting effects? am i missing something obvious already in the
> >>csound arsenal? (Basically i want to repitch & formant tweak vocals
>without
> >>"Vocoding" them...)
> >
> > Well, formant tracking and preservation is a complex business.  What
> > pvscale
> > does is quite simple (and fast): it does not shift the magnitudes along
> > with the
> > frequencies, which works for some sounds, but not for others.  Component
> > frequencies might be moved to dips in the spectrum, where they disappear,
> > so it sometimes doesn't work as well. But as I showed
> > (http://www.dafx.ca/proceedings/papers/p_275.pdf),
> > it can work well, and very cheaply. The guys at IRCAM have developed a
> > good
> > formant preservation algorithm, but it's expensive and they are very
> > snotty
> > about cheap tricks such as the one I did. If you want to read their paper,
> > it's
> > online at: http://dafx05.ssr.upm.es/Proc_DAFx05/P_030.pdf
> >
> > Victor
> >
> >
> > Victor Lazzarini
> > Music Technology Laboratory
> > Music Department
> > National University of Ireland, Maynooth
> >
> >
> >
> > Send bugs reports to this list.
> > To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> > csound"
> >
> >
>
>--
>View this message in context: 
>http://www.nabble.com/Britney-Spears-of-Destiny-%28esp.-attn-Victor-Lazzarini-%2B-others%29-tp14730945p14733136.html
>Sent from the Csound - General mailing list archive at Nabble.com.
>
>
>
>Send bugs reports to this list.
>To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe 
>csound"

Victor Lazzarini
Music Technology Laboratory
Music Department
National University of Ireland, Maynooth