• Profile picture of Ginger

Max Sound Box

Groupe Public active 1 month et 2 weeks ago

MaxSoundBox objects user group

fiddle for violin

Auteur 2 Utilisateurs souscrits |
Profile photo of Antoine ESCUDIER

I am trying to use the object fiddle in order to follow a violin. Does anyone know what arguments to use? (The default arguments 1024 1 20 3 do not work correctly for the violin).
Antoine Escudier

Août 30, 2016 à 11:40 #18921
Profile photo of arshiacont

Hi Antoine,

Here are my 2 cents after experimenting with several real-time pitch detectors few years back on Violin:

First of all, don’t use Fiddle but the later Sigmund~ object that is freely available on both Max and Puredata online. This is essentially an updated (and improved) algorithm for pitch detection and it works better especially on violin. Otherwise, concerning both Fiddle and Sigmund, I wouldn’t recommend a window size of 1024! You’d atleast need 2048 to have an acceptable spectral resolution.

On a more global note, Strings have rather unstable spectrum in contrast to winds (clarinet, sax with clear sinusoidal spectrum). This means that you’ll get the correct pitch if your input audio is clean but you might get a lot of garbage aside especially during attacks and release of a note! You might get a more stable result if you use a time-domain algorithm such as Yin~ which is part of the Max Sound Box of Ircam Forum.

I used to test and use all these in earlier versions of Antescofo score following object until I decided to do spectral computation inside to gain much more stable results! If you know the score you’d be playing in advance, I highly recommend using Antescofo~ which has been widely tested on strings and all their extended techniques during concerts and such.

Good luck,


Août 30, 2016 à 11:58 #18925
Profile photo of Antoine ESCUDIER

Hello Arshia,
Thanks for your answer. I found sigmund.
Concerning antescofo, I am planning to take the ircam training this year (despite the drastic cost increase of all ircam trainings) but doubt that I can make it work without that training…
Thanks again.
Antoine Escudier

Août 30, 2016 à 12:46 #18931
Profile photo of arshiacont

Don’t forget also to test with YIN~ object! I don’t quite remember since this was all around 2010… . I have a feeling that Yin~ could work better on signals with a lot of temporal variations such as voice and strings.

As for Antescofo, try first by looking at the embedded Max/Pd tutorials and the long online PDF documentation. If you are a Max/Pd user and familiar with Mixed Music setups learning can occur fast and without further need for training. Ofcourse, you can always send us questions in the Antescofo Forum.

Août 30, 2016 à 14:13 #18932

Vous devez être connecté pour répondre à ce sujet.

Log in now