< Back to IRCAM Forum

Issues with 'onset with descr' example


Giving a look at the ‘onset-with-descr-then-edit-and-resegment’ patcher (in the Mubu-howto-segmentation example), I came across two issues.

  1. Following the steps in the patcher, when banging the mubu.process (step 3), the first marker found is at 281.7461 ms. Bangin it again, mubu.process then find another initial marker (at 16.48526 ms).

Any reason that initial marker is missed/skipped in the first processing ?

  1. There is something that escapes me regarding the ‘chop’ process.

Using imubu ui, if I move the first marker, I see that some values change for both first and second marker stats.

Is this expected ?

I am also not totally sure how stats are run. Does the chop process compute stats using all frames between two consecutive markers or does it do so using all frames within a marker duration ?

Thanks in advance.


Congrats, Julien, you’re the 2nd one (shortly after me) to have found the missing init of onseg, already fixed for the next release. Until then, the second run gives the correct value.

  1. I don’t see that behaviour, are you sure you’re really moving the frist marker, which might be hidden behind the cursor or the left border? (Open the mubu.track for seg in matrix view to check.)

Onseg calculates stats between markers, ignoring the Duration column. This will be improved in a new way to do segmentation and stats we’re working on right now.


Hi Diemo,

Great ! Thanks for the confirmation.

As for #2, my confusion here. I was actually testing with moving second marker (index #1). But then it makes sense if stats are ran between markers. Moving a marker will also change stats for previous segment.