Gizmo~ / minimizing latency


    Jul 16 2007 | 1:53 pm

    • Jul 16 2007 | 2:25 pm
      The latency you get depends on the frame size of the pfft object. Smaller means less latency. How small will work for you depends on the amount of pitch shift you need gizmo~ to do without artifacts.
      Furthermore the latency depends on the DSP settings -> I/O vector size. Again, smaller means less latency, up 'til the size where your audio output begins to distort.
      Note that on a pc laptop you'll most likely get much latency if you don't use an external soundcard with ASIO support such as the M-Audio FW410 .
      Hth, Mattijs
      Quote: tobiasreber wrote on Mon, 16 July 2007 15:53 ---------------------------------------------------- > Hi everyone, > > I am working on a patch for a scientific research in psychiatry. In the > patch I'm using the gizmo~ object for pitch shifting. No ther signal > processing is done. > > The thing is, the pitch shifting has to be performed online and it is > measured how a person speaking into a microphone is reacting to hearing her > altered voice (the pitch shifts only have a duration of 50 or 100ms). > However, I'm getting way too much latency on the headphone/control signal to > get accurate results. > > My question is thus: How can I reduce latency in Max/MSP, or specifically > when using the gizmo~ object (which arguments are best, which signal vector > and DSP settings.) > > This is rather urgent, so thanks for any quick reply. Sorry if this has > already been discussed recently, I don't have the time to browse the > archives at the moment. > > Tobias Reber > > PS: I'm working on a Mac with OSX, 1.5Ghz, but the study will be performed > on a PC laptop (no further specifications known at the moment). > > > > ----------------------------------------------------
    • Jul 16 2007 | 2:46 pm
      Thanks a lot, Mattijs!
      I have now by trial and error come to the same realisation. I think a framesize of 1024 will work fine (it was 4096 before - I'd never worked with gizmo before and have just imitated the help file ... :)
      They study will be performed with a M-Audio FastTrack Interface. Its not FireWire like the 410 (which I use at home), but I guess it'll work.
      Thanks again,
      Tobias
      Am 16.7.2007 16:25 Uhr schrieb "Mattijs Kneppers" unter :
      > > The latency you get depends on the frame size of the pfft object. Smaller > means less latency. How small will work for you depends on the amount of pitch > shift you need gizmo~ to do without artifacts. > > Furthermore the latency depends on the DSP settings -> I/O vector size. Again, > smaller means less latency, up 'til the size where your audio output begins to > distort. > > Note that on a pc laptop you'll most likely get much latency if you don't use > an external soundcard with ASIO support such as the M-Audio FW410 . > > Hth, > Mattijs > > Quote: tobiasreber wrote on Mon, 16 July 2007 15:53 > ---------------------------------------------------- >> Hi everyone, >> >> I am working on a patch for a scientific research in psychiatry. In the >> patch I'm using the gizmo~ object for pitch shifting. No ther signal >> processing is done. >> >> The thing is, the pitch shifting has to be performed online and it is >> measured how a person speaking into a microphone is reacting to hearing her >> altered voice (the pitch shifts only have a duration of 50 or 100ms). >> However, I'm getting way too much latency on the headphone/control signal to >> get accurate results. >> >> My question is thus: How can I reduce latency in Max/MSP, or specifically >> when using the gizmo~ object (which arguments are best, which signal vector >> and DSP settings.) >> >> This is rather urgent, so thanks for any quick reply. Sorry if this has >> already been discussed recently, I don't have the time to browse the >> archives at the moment. >> >> Tobias Reber >> >> PS: I'm working on a Mac with OSX, 1.5Ghz, but the study will be performed >> on a PC laptop (no further specifications known at the moment). >> >> >> >> > ---------------------------------------------------- > > > -- > SmadSteck - http://www.smadsteck.nl > Hard- and software for interactive audiovisual sampling