Hi guys,
I’ve noticed that our matte XMesh worldspace caches don’t produce motionblur in Krakatoa, and the do produce it in VRay. Am I missing something, or we need to use objectspace caches?
Thanks!
This is interesting. I suspect somewhere along the lines the mesh is losing its vertex velocites (subframe motion).
In your scene, can you try setting the timeslider time to a non-whole-number of a frame, and see if the mesh is different than at the whole frame?
-For example, set the time slider to frame 0, then to 0.25, then to 0.5, 0.75, 1.
Was there any motion of the mesh between from 0 and 0.25?
Or 0.5 and 0.75?
Krakatoa samples at subframes to determine vertex velocity for motion blur.
Both XMeshLoaders (WS and OS) are playng smoothly in the ticks between the frames, no difference.
The real bad news is that when I cache TP, even objectspace caching doesn’t help - no mbur. TP is doing everything in WS, as far as I know.
“>Sub-Sample Geometry” is checked, right?
Ah! Checking it fixes the mblur! Makes sense 
Awesome, thanks
Here is the long story:
When Krakatoa was in use at Frantic Films, several of our plugins (for example our FLOOD meshes) caused severe slowdowns due to unnecessary updates on every subsample. So we made the decision to only enable transformation motion blur (Position/Rotation/Scale), and make Deformation motion blur optional. On top of that, we added the ability to override the number of samples actually taken from the geometry. So for example, if you had 64 passes particle motion blur, the deforming Matte Geometry could be sampled only 8 or 16 times as it would be close enough, esp. for quick tests.
This is why the option is not checked by default. That being said, you can check it, then right-click it and set it as Startup Default, so in the future it will always start up as checked.
We are considering changing the factory default to checked, too.
It’s very flexible to have these uncoupled, great stuff.