This example uses 30 channels to create a roomful of Barack Obamas.
This project is under construction and not yet available:
temporary files
max-projects/granular-timestretch
- mc_granular2.maxpat
- obama.aiff
notes:
The original idea was to adapt Andy Farnell’s PSOLA granular timestretch algorithm to run using multiple voices – using Max mc object wrapper. After many attempts, the results just sounded like mostly noise.
So I started over, using a mc granular example that comes with Max8. In this example, a metro object triggers each new voice instance. I found that by altering the rate of that metro using a [drunk] object, ie., random-walk. It would give a good approximation of the natural cycles of density that happen in a crowd of people talking.
The other changes were adding wider ranges to the sliders and more control over the overall playback speed.
Also there was a problem in the example that prevented the voices from being recycled properly.
todo:
clean up the patch, add presets, and document.