< Back to IRCAM Forum

Spat5.virtualspeakers~ alternative for better CPU performance?

Spat
#1

I have noticed that the spat5.virtualspeakers~ object is very heavy on the CPU. But I do like all the features it has. Including that I can keep using the spread inside [spat5.spat~] and then head the spreading trough the [spat5.virtualspeakers] object. Binaural in spat5.spat~ does not support spread.

Is there any way to make this object, (or an alternative solution to using spread in binaural) run lighter on the CPU?

Many Thanks, Marijn.

#2

spat5.virtualspeakers~ is somewhat CPU intensive, yes, as it uses binaural rendering.
The CPU cost depends on 1) the number of virtual speakers and 2) the HRTF that you use.
I recommend that you use SOS filters, and the lower the filter order the lower the CPU cost.
With spat5.sofa.loader, you have access to SOS filters of order 12, which should be fairly CPU efficient.

There is currently no spread factor when doing binaural rendering.
Some ideas to “spread” a source, you could :

  1. create multiple replicas of that source (and dispatch the replicas in space) or
  2. adjust the “early width” parameter in [spat5.oper]
  3. although a little experimental, spat5.decorrelate~ could also be used to create uncorrelated replicas of a signal, in order to achieve a spread/blur effect…

Best,
T.

1 Like
#3

Thank you for the suggestions, I will all of them.