BrainModular

Hello
I'm working on a patch that would automatically create markers on the transients of a recording.
I've managed to extract the peaks as an array by taking the highest levels. The problem is that some peaks are weaker than other parts of the signal.
To remedy the problem, I'd need to be able to abstract the signal.
the high levels that follow the true peaks.
I've managed to tinker with this by compressing my table, but it's still a DIY job.
If anyone has any ideas for refining the process.
I'll supply my patch as is.
Thanks
JF

Hi !
Maybe this will help ?!
edit : the "abs" module will do the trick as well

I am working on the same kind of thing, but the transients are detected while recording

check the file : Cheers

Olivar

Thaks i'll check it out !
detect while recording is maybe a simpler solution but it need to be up to date on each recording generation that could be a problem when we use overdub or circular mode.

fter verification, switching to absolute values changes nothing

joffo78 wrote: ↑

20 Sep 2023, 13:41

after verification, switching to absolute values changes nothing

the problem comes from the display array which is only 512 in size.

what if the markers where recorded on the sampler extra audio track ? their position would be in sync, even in case of circular buffer recording ?

this will be useful to get the transient position while playing (for example if your planning to use the transient infos to control speed while time-stretching (that's my plan)

otherwise, my patch can be modified to remove the existing markers that are passed by the "recording head" when in circular mode.

Cheers

Olivar

It seems to me that the size of the array is adjusted as the size of the recording increases.
Waveform detection also offers the possibility of loading several sounds and instantly obtaining the transients.

joffo78 wrote: ↑

20 Sep 2023, 17:34

It seems to me that the size of the array is adjusted as the size of the recording increases.
Waveform detection also offers the possibility of loading several sounds and instantly obtaining the transients.

You're right, but the size of the array is smaller than the size of the wavefile so you may miss some transients.

another idea would be to use this array and perform a peak analysis inside a "procedural" module (HH6 only)
https://support.ircam.fr/docs/AudioScul ... ethod.html

not simple

Hello,
I did some homework, not perfect, works only with HH6

Interesting
I'll check it soon !