Hey everyone,

I'm deep into a project involving a high-poly anatomical model composed of multiple components like bones, muscles, veins, and more. Each of these is imported as an Alembic and tagged with attributes. I'm working in Houdini 19 and using a custom Point Deform HDA to reshape these anatomical elements to fit a within a scanned "skin."
While the overall deformation is looking good, I've run into a hiccup: individual polygon islands within these anatomical components (e.g., a thumb bone, calf muscle) are starting to separate from their neighboring structures. This becomes noticeable in close-ups during animation.

I've tried a few approaches to fix this:
- Tetrahedralizing the guiding meshes for a more volumetric approach. Certainly helps a great deal.
- Using 'Pack' SOP to treat all geometry as a single unit, which only made the mesh respond at its centroid.
- Considered using 'Fuse,' but I'm wary of creating dirty geometry.
- Thought about converting each layer (bones, muscles, etc.) to watertight meshes via VDB, but this would dramatically hike up the polycount.

The goal is to make these anatomical elements deform cohesively, with absolutely minimal separations or intersections that could break the illusion in close-ups. Vellum is not an option due to the high polycount.

Any insights or suggestions on how to tackle this would be greatly appreciated. Thanks in advance!

When you say "...anatomical components are starting to separate..." are you still referring to the refitting process on the initial static model? Or are you point-deforming using an animated target mesh?

...ie is your initial refit used to generate a static output => which is then used as the source mesh in your animated deform? Or are you doing the refit and animation all in one step?


In some cases you may get better results by using two separate PointDeform SOPs. One to do *only* the Capture step, and the other to do only the Deform.

Another trick you might use would be to use the outer skin to deform only the skeletal bones. And if that's looking acceptable, then combine your skin and bones, and use them together as your deformation mesh. This would sandwich all the other anatomical elements inbetween. -sort of a simpler, stripped-down volumetric approach.

Hey John

Thanks for your insight

The seperations occur prior to animating, so yes, right now it's static, as I "squish" the anatomy from one model to another.

I'm trying to emulate the Ziva Dynamics Anatomy Transfer process which uses a tet mesh to create volumetric points to help with the warp

I've never tried using the capture/deform process! I'll look into this.

Your final note about using the skeletal bones as well as the skin is an interesting one. I didnt consider this so will give it a shot, but likely combine it with the tet mesh too at some point.

I'm 90% there, I'm just being picky and would love to nail the process for future CFX jobs.