To give you a taster of what is possible with the new array operations, I’ve prepared an
example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
(this is now one of the examples in the manual).

===========================
ksmps = 64

instr 1

 ihopsize = 256  ; hopsize
 ifftsize = 2048 ; FFT size 
 iolaps = ifftsize/ihopsize ; overlaps
 ibw = sr/ifftsize ; bin bandwidth
 kcnt init 0    ; counting vars
 krow init 0

 kOla[] init ifftsize ; overlap-add buffer
 kIn[] init ifftsize  ; input buffer
 kOlph[] init ifftsize/2+1 ; old phases
 kPhs[] init ifftsize/2+1 ; synthesis phases
 kDeltaOut[] init ifftsize/2+1 ; synthesis freqs
 kMagsOut[] init ifftsize/2+1 ; synthesis mags
 kOut[][] init iolaps, ifftsize ; output buffers

 a1 diskin2 "fox.wav",1,0,1 ; audio input

 /* every hopsize samples */
 if kcnt == ihopsize then  
   /* window and take FFT */
   kWin[] window kIn,krow*ihopsize
   kSpec[] rfft kWin
   
  /* mags & phases */
  kMags[] mags kSpec 
  kPha[] phs kSpec
  
  /* phase diffs */
  kDelta[] = kPha - kOlph 
  kOlph = kPha
  
  /* pitch shift */
  ki = 0
  iscal = 1.25 
  until ki == ifftsize/2 do
   if ki*iscal < ifftsize/2 then
     kDeltaOut[ki*iscal] = kDelta[ki]*iscal
     kMagsOut[ki*iscal] = kMags[ki]
   endif
    ki += 1
  od
  
  /* accum phases */
   kPhs = kDeltaOut + kPhs 
   kPhs unwrap kPhs
   kSpec pol2rect kMagsOut,kPhs
   
   /* IFFT + window */
   kRow[] irfft kSpec
   kWin window kRow, krow*ihopsize
   /* place it on out buffer */
   kOut setrow kWin, krow

   /* zero the ola buffer */
   kOla = 0
   /* overlap-add */
   ki = 0
   until ki == iolaps do
     kRow getrow kOut, ki
     kOla = kOla + kRow
     ki += 1
   od
  
  /* update counters */ 
  krow = (krow+1)%iolaps
  kcnt = 0
 endif

 /* shift audio in/out of buffers */
 kIn shiftin a1
 a2 shiftout kOla
 a1 delay a1, (ifftsize+ihopsize)/sr
    out a1*0.3+a2/iolaps

 /* increment counter */
 kcnt += ksmps

endin






========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie

On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
> To give you a taster of what is possible with the new array operations, I’ve prepared an
> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
> (this is now one of the examples in the manual).
>
hi,
looks interesting, is the same principle available for the stretching? 
I'm writing a small (getting now also pthreads involved, as I'm trying 
real-time) program, that records video, freezes the frame for a 
predefined time. I've got OSC-message sent to csound, when the freeze 
starts, and the basic logic in csound seems to work, sound output is 
just from diskin2 at the moment.

It would be nice to freeze sound from live input, in a way pvoc-stretch 
does. I've read Iain McCurdy's examples using granular synthesis, may 
work fine, but actually I really don't know, what would be the best method.

Any hints?

tnx

-m

To freeze the input (I think, but have not tested it), it should be the case of just bypassing the shiftin line.

========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 17:42, Matti Koskinen  wrote:

> 
> On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
>> To give you a taster of what is possible with the new array operations, I’ve prepared an
>> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
>> (this is now one of the examples in the manual).
>> 
> hi,
> looks interesting, is the same principle available for the stretching? I'm writing a small (getting now also pthreads involved, as I'm trying real-time) program, that records video, freezes the frame for a predefined time. I've got OSC-message sent to csound, when the freeze starts, and the basic logic in csound seems to work, sound output is just from diskin2 at the moment.
> 
> It would be nice to freeze sound from live input, in a way pvoc-stretch does. I've read Iain McCurdy's examples using granular synthesis, may work fine, but actually I really don't know, what would be the best method.
> 
> Any hints?
> 
> tnx
> 
> -m
> 
> 
> 
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
> 
> 
> 

Not quite so. We need to stop the new PV frames. Here’s an example



-d -o dac


/* ksmps needs to be an integer div of
   hopsize */
ksmps = 64
nchnls = 2

instr 1

 ihopsize = 256  ; hopsize
 ifftsize = 2048 ; FFT size 
 iolaps = ifftsize/ihopsize ; overlaps
 ibw = sr/ifftsize ; bin bandwidth
 kcnt init 0    ; counting vars
 krow init 0

 kOla[] init ifftsize ; overlap-add buffer
 kIn[] init ifftsize  ; input buffer
 kOlph[] init ifftsize/2+1 ; old phases
 kPhs[] init ifftsize/2+1 ; synthesis phases
 kDeltaOut[] init ifftsize/2+1 ; synthesis freqs
 kMagsOut[] init ifftsize/2+1 ; synthesis mags
 kOut[][] init iolaps, ifftsize ; output buffers

 a1 diskin2 "fox.wav",1,0,1 ; audio input

 kfreez oscil 1, 0.5, 1
 /* every hopsize samples */
 if kcnt == ihopsize then  
   /* window and take FFT */
   if kfreez == 0 then
    kWin[] window kIn,krow*ihopsize
    kSpec[] rfft kWin  
    /* mags & phases */
    kMags[] mags kSpec 
    kPha[] phs kSpec
  
    /* phase diffs */
    kDelta[] = kPha - kOlph 
    kOlph = kPha
   endif
  
  /* pitch shift */
  ki = 0
  iscal = 1
  until ki == ifftsize/2 do
   if ki*iscal < ifftsize/2 then
     kDeltaOut[ki*iscal] = kDelta[ki]*iscal
     kMagsOut[ki*iscal] = kMags[ki]
   endif
    ki += 1
  od
  
  /* accum phases */
   kPhs = kDeltaOut + kPhs 
   kPhs unwrap kPhs
   kSpec pol2rect kMagsOut,kPhs
   
   /* IFFT + window */
   kRow[] irfft kSpec
   kWin window kRow, krow*ihopsize
   /* place it on out buffer */
   kOut setrow kWin, krow

   /* zero the ola buffer */
   kOla = 0
   /* overlap-add */
   ki = 0
   until ki == iolaps do
     kRow getrow kOut, ki
     kOla = kOla + kRow
     ki += 1
   od
  
  /* update counters */ 
  krow = (krow+1)%iolaps
  kcnt = 0
 endif


 /* shift audio in/out of buffers */
  kIn shiftin a1

 a2 shiftout kOla
 a1 delay a1, (ifftsize+ihopsize)/sr
    out a2/iolaps, a2/iolaps

 /* increment counter */
 kcnt += ksmps

endin



f1 0 1024 7 0 512 0 1 1 512 1
i1 0 10





========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 18:01, Victor Lazzarini  wrote:

> To freeze the input (I think, but have not tested it), it should be the case of just bypassing the shiftin line.
> 
> ========================
> Dr Victor Lazzarini
> Senior Lecturer
> NUI Maynooth, Ireland
> victor dot lazzarini at nuim dot ie
> 
> 
> 
> 
> On 24 May 2014, at 17:42, Matti Koskinen  wrote:
> 
>> 
>> On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
>>> To give you a taster of what is possible with the new array operations, I’ve prepared an
>>> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
>>> (this is now one of the examples in the manual).
>>> 
>> hi,
>> looks interesting, is the same principle available for the stretching? I'm writing a small (getting now also pthreads involved, as I'm trying real-time) program, that records video, freezes the frame for a predefined time. I've got OSC-message sent to csound, when the freeze starts, and the basic logic in csound seems to work, sound output is just from diskin2 at the moment.
>> 
>> It would be nice to freeze sound from live input, in a way pvoc-stretch does. I've read Iain McCurdy's examples using granular synthesis, may work fine, but actually I really don't know, what would be the best method.
>> 
>> Any hints?
>> 
>> tnx
>> 
>> -m
>> 
>> 
>> 
>> Send bugs reports to
>>      https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>> 
>> 
>> 
> 
> 
> 
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
> 
> 
> 

and here’s a modification for timescaling (compressing or scaling). Probably not as good quality
as mincer/temposcal and other fsig methods; but usable I think, within limits

ksmps = 64
nchnls = 2

instr 1

 ihopsize = 256  ; hopsize
 ifftsize = 2048 ; FFT size 
 iolaps = ifftsize/ihopsize ; overlaps
 ibw = sr/ifftsize ; bin bandwidth
 kcnt init 0    ; counting vars
 krow init 0

 kOla[] init ifftsize ; overlap-add buffer
 kIn[] init ifftsize  ; input buffer
 kOlph[] init ifftsize/2+1 ; old phases
 kPhs[] init ifftsize/2+1 ; synthesis phases
 kDeltaOut[] init ifftsize/2+1 ; synthesis freqs
 kMagsOut[] init ifftsize/2+1 ; synthesis mags
 kOut[][] init iolaps, ifftsize ; output buffers

 ktscal = 1.25  ;timescale (slower < 1 < faster)
 a1 diskin2 "fox.wav",1/ktscal,0,1 ; audio input

 kfreez = 0 ; no freezing
 /* every hopsize samples */
 if kcnt == ihopsize then  
   /* window and take FFT */
   if kfreez == 0 then
    kWin[] window kIn,krow*ihopsize
    kSpec[] rfft kWin  
    /* mags & phases */
    kMags[] mags kSpec 
    kPha[] phs kSpec
  
    /* phase diffs */
    kDelta[] = kPha - kOlph 
    kOlph = kPha
   endif
  
  /* pitch shift */
  ki = 0
  until ki == ifftsize/2 do
   if ki*ktscal < ifftsize/2 then
     kDeltaOut[ki*ktscal] = kDelta[ki]*ktscal
     kMagsOut[ki*ktscal] = kMags[ki]
   endif
    ki += 1
  od
  
  /* accum phases */
   kPhs = kDeltaOut + kPhs 
   kPhs unwrap kPhs
   kSpec pol2rect kMagsOut,kPhs
   
   /* IFFT + window */
   kRow[] irfft kSpec
   kWin window kRow, krow*ihopsize
   /* place it on out buffer */
   kOut setrow kWin, krow

   /* zero the ola buffer */
   kOla = 0
   /* overlap-add */
   ki = 0
   until ki == iolaps do
     kRow getrow kOut, ki
     kOla = kOla + kRow
     ki += 1
   od
  
  /* update counters */ 
  krow = (krow+1)%iolaps
  kcnt = 0
 endif


 /* shift audio in/out of buffers */
  kIn shiftin a1

 a2 shiftout kOla
 a1 delay a1, (ifftsize+ihopsize)/sr
    out a2/iolaps, a2/iolaps

 /* increment counter */
 kcnt += ksmps

endin


========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 18:01, Victor Lazzarini  wrote:

> To freeze the input (I think, but have not tested it), it should be the case of just bypassing the shiftin line.
> 
> ========================
> Dr Victor Lazzarini
> Senior Lecturer
> NUI Maynooth, Ireland
> victor dot lazzarini at nuim dot ie
> 
> 
> 
> 
> On 24 May 2014, at 17:42, Matti Koskinen  wrote:
> 
>> 
>> On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
>>> To give you a taster of what is possible with the new array operations, I’ve prepared an
>>> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
>>> (this is now one of the examples in the manual).
>>> 
>> hi,
>> looks interesting, is the same principle available for the stretching? I'm writing a small (getting now also pthreads involved, as I'm trying real-time) program, that records video, freezes the frame for a predefined time. I've got OSC-message sent to csound, when the freeze starts, and the basic logic in csound seems to work, sound output is just from diskin2 at the moment.
>> 
>> It would be nice to freeze sound from live input, in a way pvoc-stretch does. I've read Iain McCurdy's examples using granular synthesis, may work fine, but actually I really don't know, what would be the best method.
>> 
>> Any hints?
>> 
>> tnx
>> 
>> -m
>> 
>> 
>> 
>> Send bugs reports to
>>      https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>> 
>> 
>> 
> 
> 
> 
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
> 
> 
> 

and finally, a more usual timescaling example, taking audio from a function table

 

-d -o dac


/* ksmps needs to be an integer div of
   hopsize */
ksmps = 64
nchnls = 2
0dbfs = 1

instr 1

 ihopsize = 256 ; hopsize
 ifftsize = 2048 ; FFT size 
 iolaps = ifftsize/ihopsize ; overlaps
 ibw = sr/ifftsize ; bin bandwidth
 kcnt init 0    ; counting vars
 krow init 0
 kpos init 0

 kOla[] init ifftsize ; overlap-add buffer
 kIn[] init ifftsize  ; input buffer
 kPhs[] init ifftsize/2+1 ; synthesis phases
 kOut[][] init iolaps, ifftsize ; output buffers

 ktscal = .75 ;timescale (slower < 1 < faster)

 /* every hopsize samples */
 if kcnt == ihopsize then  
     
    ki = 0
    koff = krow*ihopsize
    until ki == ifftsize do
      kIn[(ki+koff)%ifftsize] table ki+kpos,1
      ki += 1
    od
 
   /* window and take FFT */
    kWin[] window kIn, koff
    kSpec[] rfft kWin  
    /* phases */
    kPha1[] phs kSpec
    
    ki = 0
    koff = ((krow+1)%iolaps)*ihopsize
    until ki == ifftsize do
      kIn[(ki+koff)%ifftsize] table ki+kpos+ihopsize,1
      ki += 1
    od
 
   /* window and take FFT */
    kWin[] window kIn, koff
    kSpec[] rfft kWin 
     
    /* mags & phases */
    kMags[] mags kSpec 
    kPha2[] phs kSpec
    
    /* phase diffs */
    kDelta[] = kPha2 - kPha1
  
  /* accum phases */
   kPhs = kDelta + kPhs 
   kPhs unwrap kPhs
   kSpec pol2rect kMags,kPhs
   
   /* IFFT + window */
   kRow[] irfft kSpec
   kWin window kRow, krow*ihopsize
   /* place it on out buffer */
   kOut setrow kWin, krow

   /* zero the ola buffer */
   kOla = 0
   /* overlap-add */
   ki = 0
   until ki == iolaps do
     kRow getrow kOut, ki
     kOla = kOla + kRow
     ki += 1
   od
  
  /* update counters */ 
  krow = (krow+1)%iolaps
  kcnt = 0
  kpos += ihopsize*ktscal
 endif


 /* shift audio out of buffers */
 a2 shiftout kOla
    out a2/iolaps, a2/iolaps

 /* increment counter */
 kcnt += ksmps

endin



f1 0 0 1 "fox.wav" 0 0 1
i1 0 60




========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 18:25, Victor Lazzarini  wrote:

> and here’s a modification for timescaling (compressing or scaling). Probably not as good quality
> as mincer/temposcal and other fsig methods; but usable I think, within limits
> 
> ksmps = 64
> nchnls = 2
> 
> instr 1
> 
> ihopsize = 256  ; hopsize
> ifftsize = 2048 ; FFT size 
> iolaps = ifftsize/ihopsize ; overlaps
> ibw = sr/ifftsize ; bin bandwidth
> kcnt init 0    ; counting vars
> krow init 0
> 
> kOla[] init ifftsize ; overlap-add buffer
> kIn[] init ifftsize  ; input buffer
> kOlph[] init ifftsize/2+1 ; old phases
> kPhs[] init ifftsize/2+1 ; synthesis phases
> kDeltaOut[] init ifftsize/2+1 ; synthesis freqs
> kMagsOut[] init ifftsize/2+1 ; synthesis mags
> kOut[][] init iolaps, ifftsize ; output buffers
> 
> ktscal = 1.25  ;timescale (slower < 1 < faster)
> a1 diskin2 "fox.wav",1/ktscal,0,1 ; audio input
> 
> kfreez = 0 ; no freezing
> /* every hopsize samples */
> if kcnt == ihopsize then  
>   /* window and take FFT */
>   if kfreez == 0 then
>    kWin[] window kIn,krow*ihopsize
>    kSpec[] rfft kWin  
>    /* mags & phases */
>    kMags[] mags kSpec 
>    kPha[] phs kSpec
> 
>    /* phase diffs */
>    kDelta[] = kPha - kOlph 
>    kOlph = kPha
>   endif
> 
>  /* pitch shift */
>  ki = 0
>  until ki == ifftsize/2 do
>   if ki*ktscal < ifftsize/2 then
>     kDeltaOut[ki*ktscal] = kDelta[ki]*ktscal
>     kMagsOut[ki*ktscal] = kMags[ki]
>   endif
>    ki += 1
>  od
> 
>  /* accum phases */
>   kPhs = kDeltaOut + kPhs 
>   kPhs unwrap kPhs
>   kSpec pol2rect kMagsOut,kPhs
> 
>   /* IFFT + window */
>   kRow[] irfft kSpec
>   kWin window kRow, krow*ihopsize
>   /* place it on out buffer */
>   kOut setrow kWin, krow
> 
>   /* zero the ola buffer */
>   kOla = 0
>   /* overlap-add */
>   ki = 0
>   until ki == iolaps do
>     kRow getrow kOut, ki
>     kOla = kOla + kRow
>     ki += 1
>   od
> 
>  /* update counters */ 
>  krow = (krow+1)%iolaps
>  kcnt = 0
> endif
> 
> 
> /* shift audio in/out of buffers */
>  kIn shiftin a1
> 
> a2 shiftout kOla
> a1 delay a1, (ifftsize+ihopsize)/sr
>    out a2/iolaps, a2/iolaps
> 
> /* increment counter */
> kcnt += ksmps
> 
> endin
> 
> 
> ========================
> Dr Victor Lazzarini
> Senior Lecturer
> NUI Maynooth, Ireland
> victor dot lazzarini at nuim dot ie
> 
> 
> 
> 
> On 24 May 2014, at 18:01, Victor Lazzarini  wrote:
> 
>> To freeze the input (I think, but have not tested it), it should be the case of just bypassing the shiftin line.
>> 
>> ========================
>> Dr Victor Lazzarini
>> Senior Lecturer
>> NUI Maynooth, Ireland
>> victor dot lazzarini at nuim dot ie
>> 
>> 
>> 
>> 
>> On 24 May 2014, at 17:42, Matti Koskinen  wrote:
>> 
>>> 
>>> On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
>>>> To give you a taster of what is possible with the new array operations, I’ve prepared an
>>>> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
>>>> (this is now one of the examples in the manual).
>>>> 
>>> hi,
>>> looks interesting, is the same principle available for the stretching? I'm writing a small (getting now also pthreads involved, as I'm trying real-time) program, that records video, freezes the frame for a predefined time. I've got OSC-message sent to csound, when the freeze starts, and the basic logic in csound seems to work, sound output is just from diskin2 at the moment.
>>> 
>>> It would be nice to freeze sound from live input, in a way pvoc-stretch does. I've read Iain McCurdy's examples using granular synthesis, may work fine, but actually I really don't know, what would be the best method.
>>> 
>>> Any hints?
>>> 
>>> tnx
>>> 
>>> -m
>>> 
>>> 
>>> 
>>> Send bugs reports to
>>>     https://github.com/csound/csound/issues
>>> Discussions of bugs and features can be posted here
>>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>>> 
>>> 
>>> 
>> 
>> 
>> 
>> Send bugs reports to
>>       https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>> 
>> 
>> 
> 
> 
> 
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
> 
> 
> 

Thank you very much!
I don't get any sound though
tnx

-m


On 05/24/2014 08:25 PM, Victor Lazzarini wrote:
> and here’s a modification for timescaling (compressing or scaling). Probably not as good quality
> as mincer/temposcal and other fsig methods; but usable I think, within limits
>
> ksmps = 64
> nchnls = 2
>
> instr 1
>
>   ihopsize = 256  ; hopsize
>   ifftsize = 2048 ; FFT size
>   iolaps = ifftsize/ihopsize ; overlaps
>   ibw = sr/ifftsize ; bin bandwidth
>   kcnt init 0    ; counting vars
>   krow init 0
>
>   kOla[] init ifftsize ; overlap-add buffer
>   kIn[] init ifftsize  ; input buffer
>   kOlph[] init ifftsize/2+1 ; old phases
>   kPhs[] init ifftsize/2+1 ; synthesis phases
>   kDeltaOut[] init ifftsize/2+1 ; synthesis freqs
>   kMagsOut[] init ifftsize/2+1 ; synthesis mags
>   kOut[][] init iolaps, ifftsize ; output buffers
>
>   ktscal = 1.25  ;timescale (slower < 1 < faster)
>   a1 diskin2 "fox.wav",1/ktscal,0,1 ; audio input
>
>   kfreez = 0 ; no freezing
>   /* every hopsize samples */
>   if kcnt == ihopsize then
>     /* window and take FFT */
>     if kfreez == 0 then
>      kWin[] window kIn,krow*ihopsize
this part is bit unclear. Does the kIn shiftin a1 in the end of the 
instr fill the buffer after enough loops?
>      kSpec[] rfft kWin
>      /* mags & phases */
>      kMags[] mags kSpec
>      kPha[] phs kSpec
>    
>      /* phase diffs */
>      kDelta[] = kPha - kOlph
>      kOlph = kPha
>     endif
>    
>    /* pitch shift */
>    ki = 0
>    until ki == ifftsize/2 do
>     if ki*ktscal < ifftsize/2 then
>       kDeltaOut[ki*ktscal] = kDelta[ki]*ktscal
>       kMagsOut[ki*ktscal] = kMags[ki]
>     endif
>      ki += 1
>    od
>    
>    /* accum phases */
>     kPhs = kDeltaOut + kPhs
>     kPhs unwrap kPhs
>     kSpec pol2rect kMagsOut,kPhs
>     
>     /* IFFT + window */
>     kRow[] irfft kSpec
>     kWin window kRow, krow*ihopsize
>     /* place it on out buffer */
>     kOut setrow kWin, krow
>
>     /* zero the ola buffer */
>     kOla = 0
>     /* overlap-add */
>     ki = 0
>     until ki == iolaps do
>       kRow getrow kOut, ki
>       kOla = kOla + kRow
>       ki += 1
>     od
>    
>    /* update counters */
>    krow = (krow+1)%iolaps
>    kcnt = 0
>   endif
>
>
>   /* shift audio in/out of buffers */
>    kIn shiftin a1
>
>   a2 shiftout kOla
>   a1 delay a1, (ifftsize+ihopsize)/sr
>      out a2/iolaps, a2/iolaps
>
>   /* increment counter */
>   kcnt += ksmps
>
> endin
>
>
> ========================
> Dr Victor Lazzarini
> Senior Lecturer
> NUI Maynooth, Ireland
> victor dot lazzarini at nuim dot ie
>
>
>
>
> On 24 May 2014, at 18:01, Victor Lazzarini  wrote:
>
>> To freeze the input (I think, but have not tested it), it should be the case of just bypassing the shiftin line.
>>
>> ========================
>> Dr Victor Lazzarini
>> Senior Lecturer
>> NUI Maynooth, Ireland
>> victor dot lazzarini at nuim dot ie
>>
>>
>>
>>
>> On 24 May 2014, at 17:42, Matti Koskinen  wrote:
>>
>>> On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
>>>> To give you a taster of what is possible with the new array operations, I’ve prepared an
>>>> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
>>>> (this is now one of the examples in the manual).
>>>>
>>> hi,
>>> looks interesting, is the same principle available for the stretching? I'm writing a small (getting now also pthreads involved, as I'm trying real-time) program, that records video, freezes the frame for a predefined time. I've got OSC-message sent to csound, when the freeze starts, and the basic logic in csound seems to work, sound output is just from diskin2 at the moment.
>>>
>>> It would be nice to freeze sound from live input, in a way pvoc-stretch does. I've read Iain McCurdy's examples using granular synthesis, may work fine, but actually I really don't know, what would be the best method.
>>>
>>> Any hints?
>>>
>>> tnx
>>>
>>> -m
>>>
>>>
>>>
>>> Send bugs reports to
>>>       https://github.com/csound/csound/issues
>>> Discussions of bugs and features can be posted here
>>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>>>
>>>
>>>
>>
>>
>> Send bugs reports to
>>         https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>>
>>
>>
>
>
> Send bugs reports to
>          https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>
>
>
>

What do you mean you don’t get any sound? It works here. Did you modify the example?
shiftin fills the buffer continuously, and every ihopsize samples we do a spectral analysis.


========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 21:34, Matti Koskinen  wrote:

> Thank you very much!
> I don't get any sound though
> tnx
> 
> -m
> 
> 
> On 05/24/2014 08:25 PM, Victor Lazzarini wrote:
>> and here’s a modification for timescaling (compressing or scaling). Probably not as good quality
>> as mincer/temposcal and other fsig methods; but usable I think, within limits
>> 
>> ksmps = 64
>> nchnls = 2
>> 
>> instr 1
>> 
>>  ihopsize = 256  ; hopsize
>>  ifftsize = 2048 ; FFT size
>>  iolaps = ifftsize/ihopsize ; overlaps
>>  ibw = sr/ifftsize ; bin bandwidth
>>  kcnt init 0    ; counting vars
>>  krow init 0
>> 
>>  kOla[] init ifftsize ; overlap-add buffer
>>  kIn[] init ifftsize  ; input buffer
>>  kOlph[] init ifftsize/2+1 ; old phases
>>  kPhs[] init ifftsize/2+1 ; synthesis phases
>>  kDeltaOut[] init ifftsize/2+1 ; synthesis freqs
>>  kMagsOut[] init ifftsize/2+1 ; synthesis mags
>>  kOut[][] init iolaps, ifftsize ; output buffers
>> 
>>  ktscal = 1.25  ;timescale (slower < 1 < faster)
>>  a1 diskin2 "fox.wav",1/ktscal,0,1 ; audio input
>> 
>>  kfreez = 0 ; no freezing
>>  /* every hopsize samples */
>>  if kcnt == ihopsize then
>>    /* window and take FFT */
>>    if kfreez == 0 then
>>     kWin[] window kIn,krow*ihopsize
> this part is bit unclear. Does the kIn shiftin a1 in the end of the instr fill the buffer after enough loops?
>>     kSpec[] rfft kWin
>>     /* mags & phases */
>>     kMags[] mags kSpec
>>     kPha[] phs kSpec
>>        /* phase diffs */
>>     kDelta[] = kPha - kOlph
>>     kOlph = kPha
>>    endif
>>      /* pitch shift */
>>   ki = 0
>>   until ki == ifftsize/2 do
>>    if ki*ktscal < ifftsize/2 then
>>      kDeltaOut[ki*ktscal] = kDelta[ki]*ktscal
>>      kMagsOut[ki*ktscal] = kMags[ki]
>>    endif
>>     ki += 1
>>   od
>>      /* accum phases */
>>    kPhs = kDeltaOut + kPhs
>>    kPhs unwrap kPhs
>>    kSpec pol2rect kMagsOut,kPhs
>>        /* IFFT + window */
>>    kRow[] irfft kSpec
>>    kWin window kRow, krow*ihopsize
>>    /* place it on out buffer */
>>    kOut setrow kWin, krow
>> 
>>    /* zero the ola buffer */
>>    kOla = 0
>>    /* overlap-add */
>>    ki = 0
>>    until ki == iolaps do
>>      kRow getrow kOut, ki
>>      kOla = kOla + kRow
>>      ki += 1
>>    od
>>      /* update counters */
>>   krow = (krow+1)%iolaps
>>   kcnt = 0
>>  endif
>> 
>> 
>>  /* shift audio in/out of buffers */
>>   kIn shiftin a1
>> 
>>  a2 shiftout kOla
>>  a1 delay a1, (ifftsize+ihopsize)/sr
>>     out a2/iolaps, a2/iolaps
>> 
>>  /* increment counter */
>>  kcnt += ksmps
>> 
>> endin
>> 
>> 
>> ========================
>> Dr Victor Lazzarini
>> Senior Lecturer
>> NUI Maynooth, Ireland
>> victor dot lazzarini at nuim dot ie
>> 
>> 
>> 
>> 
>> On 24 May 2014, at 18:01, Victor Lazzarini  wrote:
>> 
>>> To freeze the input (I think, but have not tested it), it should be the case of just bypassing the shiftin line.
>>> 
>>> ========================
>>> Dr Victor Lazzarini
>>> Senior Lecturer
>>> NUI Maynooth, Ireland
>>> victor dot lazzarini at nuim dot ie
>>> 
>>> 
>>> 
>>> 
>>> On 24 May 2014, at 17:42, Matti Koskinen  wrote:
>>> 
>>>> On 05/24/2014 12:40 PM, Victor Lazzarini wrote:
>>>>> To give you a taster of what is possible with the new array operations, I’ve prepared an
>>>>> example implementing a simple phase vocoder pitch shifter/harmoniser from basic principles
>>>>> (this is now one of the examples in the manual).
>>>>> 
>>>> hi,
>>>> looks interesting, is the same principle available for the stretching? I'm writing a small (getting now also pthreads involved, as I'm trying real-time) program, that records video, freezes the frame for a predefined time. I've got OSC-message sent to csound, when the freeze starts, and the basic logic in csound seems to work, sound output is just from diskin2 at the moment.
>>>> 
>>>> It would be nice to freeze sound from live input, in a way pvoc-stretch does. I've read Iain McCurdy's examples using granular synthesis, may work fine, but actually I really don't know, what would be the best method.
>>>> 
>>>> Any hints?
>>>> 
>>>> tnx
>>>> 
>>>> -m
>>>> 
>>>> 
>>>> 
>>>> Send bugs reports to
>>>>      https://github.com/csound/csound/issues
>>>> Discussions of bugs and features can be posted here
>>>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>>> Send bugs reports to
>>>        https://github.com/csound/csound/issues
>>> Discussions of bugs and features can be posted here
>>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>>> 
>>> 
>>> 
>> 
>> 
>> Send bugs reports to
>>         https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>> 
>> 
>> 
>> 
> 
> 
> 
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"

On 05/24/2014 11:42 PM, Victor Lazzarini wrote:
> What do you mean you don’t get any sound? It works here. Did you modify the example?
> shiftin fills the buffer continuously, and every ihopsize samples we do a spectral analysis.
>
>
in the end csound gives overall amps 0.00000 0.000000
I changed only my own mono wav-file to diskin2, nothing else
Tried both on osx 10.9 after brew install csound --HEAD and ubuntu 13.04 
after git clone csound.
ksmps is 64 *OR* I had ksmp=64, just checked, after fix, yes! sound.

sorry

-m

Good to know. Yes, ksmps=64 is required.
========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 21:51, Matti Koskinen  wrote:

> 
> On 05/24/2014 11:42 PM, Victor Lazzarini wrote:
>> What do you mean you don’t get any sound? It works here. Did you modify the example?
>> shiftin fills the buffer continuously, and every ihopsize samples we do a spectral analysis.
>> 
>> 
> in the end csound gives overall amps 0.00000 0.000000
> I changed only my own mono wav-file to diskin2, nothing else
> Tried both on osx 10.9 after brew install csound --HEAD and ubuntu 13.04 after git clone csound.
> ksmps is 64 *OR* I had ksmp=64, just checked, after fix, yes! sound.
> 
> sorry
> 
> -m
> 
> 
> 
> 
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
> 
> 
> 

osx fine, but on linux:

Score finished in csoundPerform().
inactive allocs returned to freespace
end of score.           overall amps:  0.05776  0.05776
        overall samples out of range:        0        0
0 errors in performance
Elapsed time at end of performance: real: 9.909s, CPU: 3.770s
324 2730 sample blks of 64-bit floats written to dac
*** Error in `csound': free(): invalid size: 0x00000000016d94b0 ***
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x80996)[0x7f2abb95a996]
/usr/local/lib/libcsound64.so.6.0(+0x5b638)[0x7f2abbf1a638]
/usr/local/lib/libcsound64.so.6.0(+0x1b41a6)[0x7f2abc0731a6]
/usr/local/lib/libcsound64.so.6.0(csoundDestroy+0xb1)[0x7f2abc06f0d9]
csound[0x4015c3]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf5)[0x7f2abb8fbde5]
csound[0x40118c]


-m

fixed now in git
========================
Dr Victor Lazzarini
Senior Lecturer
NUI Maynooth, Ireland
victor dot lazzarini at nuim dot ie




On 24 May 2014, at 22:10, Matti Koskinen  wrote:

> osx fine, but on linux:
> 
> Score finished in csoundPerform().
> inactive allocs returned to freespace
> end of score.           overall amps:  0.05776  0.05776
>       overall samples out of range:        0        0
> 0 errors in performance
> Elapsed time at end of performance: real: 9.909s, CPU: 3.770s
> 324 2730 sample blks of 64-bit floats written to dac
> *** Error in `csound': free(): invalid size: 0x00000000016d94b0 ***
> ======= Backtrace: =========
> /lib/x86_64-linux-gnu/libc.so.6(+0x80996)[0x7f2abb95a996]
> /usr/local/lib/libcsound64.so.6.0(+0x5b638)[0x7f2abbf1a638]
> /usr/local/lib/libcsound64.so.6.0(+0x1b41a6)[0x7f2abc0731a6]
> /usr/local/lib/libcsound64.so.6.0(csoundDestroy+0xb1)[0x7f2abc06f0d9]
> csound[0x4015c3]
> /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf5)[0x7f2abb8fbde5]
> csound[0x40118c]
> 
> 
> -m
> 
> 
> 
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
> 
> 
>