If I have data coming into it, from SOPs, with subframe motion(either from a transform live, or cached
subframes, or a timeblend)and I want to bake subframe data into the exported usd files, I have to use
a cache LOP before the GeometryClipSequence.
The GeometryClipSequence LOP has a subframe sampling section, with the same parms as the cache LOP, but it
absolutely doesn't work. You can see in the openGL or karma viewport that it's not correct in terms of timeline
and it appears to be scaling the input as well. It seems like there is some transpose matrix issue maybe?
But on closer inspection even the cache LOP is causing some scaling of the geometry.
Attached is a hip file showing this.
I have not found one piece of Documentation about workflows to write out subframe data using this new node.
Can we please get a working, concrete example of how to use it to make stitch clips that contain specified subframes/shutter
durations, and doesn't scale the incoming geometric data?
In production we would explicitly be providing the subframe data, either as cached to disk, whole samples with a timeblend, or if it were being animated directly in SOPs "live" we would just pass that along.
The other input option being already written to disk .usd or .bgeo sequence with subframes already existing, and this LOP purely making the stitched clip.
whole frame linear interp is not good enough so we routinely output enough samples to exist in the correct point in time relative to the shutter duration and geom samples in the renderer.
Cheers,
L
