nm-grainer

My stereo granular synthesis patch for PD, released to the pd email list in 2002. You can download a copy of the patch here.

nm-grainer is a variably polyphonic, synchronous or asynchronous sound-file granulator. It is set for a maximum of 30 voices, but the limit could be changed by editing a few bits of the patch. I've easily run 60 grains on a PIII 900MHz CPU. The patch is designed so that a file may be played synchronously, then a variable amount of asynchronicity may be introduced by adding randomness to each parameter.

Here's a screenshot.

nm-grainer uses pix's clever not-quite-poly abstraction to implement the real-time variable polyphony - i.e. the amount of processing used by the patch varies according to the number of grains playing simultaneously (which can be set as a parameter by the user).

Note that while nm-grainer is a fairly reasonable granular synthesis implementation, it is inherently limited in audio quality by the 64-sample control signal blocksize in PD. This means that grains will not necessarily get triggered at exactly the time intended by the grain sequencer. So there is an inherent jitter that is particularly noticable for short grain lengths. The audible artifact is a strange pitched kind of noise.

A way around that limitation would be to use the Time Tagged Triggers (T3) objects from IEMLIB by Thomas Musil. I have yet to implement a new version of the patch using the T3 objects.