Hi,

I would like to detect peaks in an audiofile and and be able to trigger the parts. Is there a fav patch out there? Just want to explore a bit.

I will be fine with just being able to get the peaks listed in an array by time of occurrence.

Perhaps [bonk~]?

@cfry @katjav has updated her Slice//Jockey for vanilla. It is made to be used with a microphone.... but it is Pd so could be modified......... https://www.katjaas.nl/slicejockey/slicejockey.html
David.

@cfry I had completely forgotten about this one......... http://www.pdpatchrepo.info/patches/patch/10
It will need some work probably to update it for Pd vanilla.
David.

@whale-av said:

@cfry @katjav has updated her Slice//Jockey for vanilla. It is made to be used with a microphone.... but it is Pd so could be modified......... https://www.katjaas.nl/slicejockey/slicejockey.html
David.

I tried to get it to work but once I had told my Mac to allow every single lib file to be used (Apple...) the slicejockey patch crash pd. However, it is fine since I do not intend to use a beat slicer like this. Maybe I can look inside to see if I find something I can use.

@alexandros said:

Perhaps [bonk~]?

What I would like to do is detect peaks in an audio file at the same time as I load it into a table. Not realtime, or just as fast as possible.

As of now I have a patch where one table holds the sound file, another one holds 12 positions in ms, (and another one holds these 12 positions in percent.)

Either I use a list with predefined positions to load with a specific sound, or I divide the positions evenly across the sound file. In this latter case I would like a function where the locations are upon load nudged to transients to make each location more useful. Later on I would like to try to select positions by spectral content also/instead.

I attach my patch where I can load the 3 arrays on the fly. To maybe add some clarity.

sound-0-wip.pd

Could [bonk~] be used to do this offline type of action?

@cfry There are several ways to get Pd to process audio offline:
https://forum.pdpatchrepo.info/topic/13125/batch-processing-audio-faster-than-realtime
https://forum.pdpatchrepo.info/topic/13525/banging-switch-performs-audio-computations-offline
https://forum.pdpatchrepo.info/topic/12788/offline-analysis-on-a-song/10

@jameslo said:

@cfry There are several ways to get Pd to process audio offline:
https://forum.pdpatchrepo.info/topic/13125/batch-processing-audio-faster-than-realtime
https://forum.pdpatchrepo.info/topic/13525/banging-switch-performs-audio-computations-offline
https://forum.pdpatchrepo.info/topic/12788/offline-analysis-on-a-song/10

Thanks. While browsing those threads I realised I can look into timbreID and there is an example patch called segment.pd that "use BFCC deltas to find sudden changes in timbre", non-realtime. It seems really solid, worked perfect while testing it out. And there is bark~ in there which also could work. It is similar to bonk~.

I may have to re-think the flow of the whole operation since I may not really need to do this during performance, when switching audio files, as long as I have analysed the files I may use beforehand, and stored the peak index arrays/lists, and linked them to respectively audio file. And then switch peak location list when I load the audio file. But the purpose is also to make this smooth while trying out what audio material to use, I need to minimise preparation steps.