Getting better pitch data from buffer slices

Heya! I’m trying to do some corpus/concatenative granular stuff similar to @rodrigo.constanzo 's c-combine device. Currently I’m at the stage of building an accurately analyzed corpus.

I’m having a bit of trouble getting a good pitch value per slice. (ampgate with onsets and offsets > bufpitch > mean stats) Listening to real-time tracking via fluid.pitch~ or sigmund~ gives me a pretty accurate tone when the notes are loud, but get whacked out when the volume goes down - certainly throwing off the mean value.

I’m thinking there’s something I can do with setting a threshold of confidence or loudness, throwing away everything below a certain value and then getting the mean of the solid stuff. Just not sure where to look.


1 Like

Alhtough I didn’t get anywhere perfect in the end, there’s some useful discussion/thinking in this older thread:

(some of the patches may not work anymore since the syntax has changed a few times since then)


Hey @jessephamel which CCE are you on?

There are good examples of massaging the statistics of pitch in the fluid.bufstats~ and FluidBufStats help files for Max and SuperCollider. This specific tab might be of interest to you if you are using Max:

1 Like