Start a new topic

ChatGPT v Orba 2

This is a continuation of the ChatGPT v Orba 1 thread. My focus is on the Orba 2 now so I thought I'd move it here.

Having made some progress converting MIDI files to Orba 1, I'm seeing if the same routines will work on Orba 2.

There seems to be some difference in the interpretation of loopData; some sequences work and some don't. For instance, here's Für Elise on Orba 2, which copied straight over from Orba 1.

Others get jumbled up though, for some reason. I just narrowed down a specific difference where the Orba 1 correctly played a simple block of note data as a five-note sequence, but the Orba 2 played it as a four-note sequence.






By experimenting with different values I eventually found that the Orba 2 interpreted it correctly if I changed the duration of the second note from 110 to 120. Anything below 120 went wrong, but 120 and above worked OK.

I took a look at the console output for Orba 1 and noticed a couple of values reported as 120; "loopBarEndTolerance" and "quantStartSnapTicks", so it may be related to one of those.

(...correction; timing, not duration...)

...yes, that seems to be the key to it; I just got Flight of the Bumblebee working on Orba 2 by stretching it so that the relative note timings were above 120. The only problem is that it's now far too slow at the maximum tempo. I need to try recording something fast on the Orba 2 so I can see how it's represented, but I haven't got as far as tweaking the presets to cut out the expression data yet, so that's the next step.

I was just taking another look at this and it seems to be a fundamental restriction or bug with the Orba 2 when recording anhd playing back fast sequences. Here's a video which demonstrates the issue, recorded at a tempo of 40. I've raised a support ticket.

This turns out to be a known issue with quantise, which gets enabled automatically even after turning it off, so I'll wait for a fix.

Login or Signup to post a comment