I'm using vellum to fill thousands of small candy objects inside a large S-shaped container.

My technique is as follows :

* Incoming candy geo has been remeshed and shrunk down to ~1/10 scale and scattered on points inside the S-Shape
* A secondary state is made at the SOP level with the geo at regular scale -- to be used as a vellum rest blend.
* A unique piece attribute has been applied to each candy to be used later
* Candy is configured as a strut softbody
* Vellum rest blend is added inside vellum solver, set to "each frame", and keyframed to transition from small to big over 48 frames or so.
* Vellum solver is run with 6 substeps or so \ 600 constraint iterations.

This seems to works great -- My small candy objects slowly scale up, collide, and fill up my container -- as depicted in the attached screenshot.

My issue is that when using point deform to take my results and apply it to my high-res object, it never fully fills the shape of the proxy ( see attached ). It's as if it's lagging behind a couple frames and never catches up. I'm specifying my piece attribute in the point deform so it evaluates each candy individually as to avoid artifacts. I've tried different radius values to no avail.

I've used this technique before successfully on sims of a much a smaller scale. I'm at a lost as to why it's not working in this scenario.

Just to be sure: Did you do the point deform with the equally scaled mesh versions? It sounds like there is an issue with an offset of these meshes which will be kept when using this node. You could also use the xyzdist function in VEX to be sure the meshes are snapped onto each other.

Yes, it's being point deformed from low-res to high-res while both are in their small state. If I A/B the inputs, the low and high are in the same position.