Initial Preparation of Cloth for Collision Simulation in Houdini

When your character includes cloth simulation, you’ll typically model separate meshes for the cloth and the body. The body acts as the cloth’s collision proxy. The degree of separation depends on how much cloth you’re simulating.

A straightforward workflow is to start in a T-pose, then animate the character while the cloth remains a dynamic object. In an ideal world this works, but more often than not it fails—especially as the animation lengthens.

A proven technique to improve this workflow is to animate the cloth with joints just like the body. This mirrors the standard approach used when animating a character without cloth simulation. Then run the cloth simulation so it follows the original animation, biasing the influence toward the method you find most favorable. This reduces unwanted entangling and makes the cloth behavior more predictable.

This approach still isn’t perfect. In some cases the initial pass (no simulation) already has cloth intersecting itself or the body, so the simulation will also collide since it follows that initial motion.

A practical mitigation is to manually push out the intersections in the cloth mesh and use that cleaned setup as the base for animation.

This setup specifically addresses that problem: push_sdf_for_cloth_collission