# Stitchclip weirdness

**URL:** <https://forum.aousd.org/t/stitchclip-weirdness/1350>\
**Category:** USD\
**Created:** [March 19, 2024, 7:27am UTC](https://forum.aousd.org/t/stitchclip-weirdness/1350 "2024-03-19T07:27:25Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![patricknagle](https://avatars.discourse-cdn.com/v4/letter/p/f17d59/32.png) [@patricknagle](https://forum.aousd.org/u/patricknagle)\
**Post date:** [March 19, 2024, 7:27am UTC](https://forum.aousd.org/t/stitchclip-weirdness/1350/1 "2024-03-19T07:27:25Z")

</div>

Hi,

We are caching some heavy basis curves that are going to be deformed in the pipeline and are running into a weird situation where the stitchclips.topology.usda seems to be _duplicating_ the very dense curves into the topology file.

I noticed that each frame of the basis curves we have cached do not contain timesampled attributes - is this a problem?

Any tips would be appreciated!

---

<div class="post-metadata">

**Author:** ![spiff](https://sea2.discourse-cdn.com/flex016/user_avatar/forum.aousd.org/spiff/32/22_2.png) [@spiff](https://forum.aousd.org/u/spiff)\
**Post date:** [March 29, 2024, 12:16am UTC](https://forum.aousd.org/t/stitchclip-weirdness/1350/2 "2024-03-29T00:16:47Z")

</div>

Hi Patrick… the first issue definitely sounds like a bug. If you can provide a repro, would you file an Issue?

It is not a “problem” that all clips do not contain data for all prims. By default, on those frames where clips-with-missing-data are active, the clipSet will provide no data and so likely the BasisCurves will disappear. But there is an arg you can provide to `usdstitchclips --interpolateMissingClipValues` that will “fill in the gaps” by interpolating (or holding from the left if arrays differ in size) between the most proximal clips in the clipSet that **do** provide data. Note [this option is a bit more expensive for value resolution](https://openusd.org/release/api/_usd __page__ value_clips.html#Usd_ValueClips_ClipValueResolution_InterpolatingGaps) since we need to search to find the relevant clips, but we find it useful in simulation cases where a single sample **is authored** in each clip, to get better motion blur because it also enables interpolation between _adjacent_ clips, i.e. just as if the clips were “flattened” into ordinary timeSamples, and in that case, the extra cost is negligible, I believe.