Fluid simulation with Zibra AI

Prototyped seawall interactions in Unity using Zibra Liquid. Kept tests in small, focused scene layers plus prefab variants to swap geometry, tune sims, and record results quickly.

Date: Nov 16, 2024

Shot setup and layering

Each test lived in a small additive scene to swap walls and water settings without touching the core XR rig. The base scene held lighting, camera, and a clean ground plane. Test scenes contained only a Zibra Liquid container, emitters, and the current seawall prefab(s).

Prefab variants: WALL_A / WALL_B / WALL_C with different structural attributes.
ScriptableObject presets for sim settings (container size, grid resolution, viscosity, surface tension).
Colliders built as Zibra SDFs and baked per prefab. Geometry was simplified for SDF generation while keeping the original structure.

Leak test: short wall segment. Target was zero seepage along seams.

Wall behavior splash test: two blocks.

Tide test: larger domain with controlled simulated tides to stress wall edge cases.

Railings and seawalls: remodel and sanity checks

Remodeled low-poly railing section dropped into the sim scene. Used as a stand-in to verify collider fidelity and splash behavior.

Seawall and railing pieces were retopologized for Zibra SDF colliders. Triangle count stayed low on the render mesh, and the SDF handled contact. Tests focused on clean contact at posts, no micro-leaks at panel joins, and consistent cresting where water meets geometry.

Simulation notes

Container: sized just large enough around the test to keep voxel count down. Shallow trays for speed, deeper basins for tide trials.
Key dials per shot: viscosity for sheet versus splash, surface tension for breakup, emitter rate for fill speed, and boundary stickiness to reduce jitter at the wall.
Validation: screen captures from both Scene and Game views for side-by-side review.

Handoff to sim and device

For each wall variant, a minimal test scene was duplicated, assigned a preset, and built to Quest. Zibra SDFs were baked once per prefab and checked into source so results matched across the team. On device, short clips were recorded for leak checks, crest height, and recovery after impact. If a wall failed (seepage or unstable splash), the prefab variant was swapped or SDF accuracy was increased and the test was re-run.

Zibra Liquid documentation