I am willing to use the extra Framelib for a special psola synthesis.
I would use fluid.bufpitch~ using the yin algorithm. Its confidence factor would allow me separating pitch synchronous granular synthesis from random granular synthesis.
I only miss the onset for each “formant”, grain, found in the auto-correlation. Would a super fast and precise fluid.buftransientslice~ do the job ? Or would I need a special fluid.bufpitch~ giving me such positions ?
Does already it exists a way to do that ?
There’s an option in
fluid.noveltyslice to segment on pitch instead of spectrum, which might work (notwithstanding the inefficiency of having to do the pitch detection twice). However, for PSOLA, you might also want to couple this with
zerox~ to try and capture whole cycles in your pitched grains?
I don’t know if the
pipo.psy object would be more use here?
Thank you !
I will give it a try but I indeed think there would maybe be a lack of precision.
I would indeed rather avoid
pipo.psy and only use Flucoma and FrameLib externals
So I will do some experiments soon.