Blender Simulation Effects

Treat physical causes as simulation problems before treating their final appearance as static modeling.

Suite Authority

Follow ../blender-production-router/references/production-protocol.md and ../blender-production-router/references/native-component-contract.md. The Router owns route, stage, budgets, rollback, and pauses. This skill owns physical participants, solver setup, preview, cache, and stability evidence only; it must not restart analysis or silently replace a simulation.

System Choice And Asset Boundary

Read ../blender-production-router/references/system-choice-contract.md. This Skill owns physical causality, participants, solver choice, cache, and stability. Geometry Nodes may generate emitters, instances, masks, or post-process a result when it gives a documented procedural advantage, but a hybrid must declare one time_state_owner; a graph cannot silently replace gravity, collision, flow, or material response. A requested local particle or VFX asset is discovered read-only by blender-local-asset-library; this Skill owns compatibility, integration, and simulation evidence.

Upstream Gate

Require minimum Production Analysis for a reversible low-resolution test and an approved Part Graph, real scale, classified physical participants, and passed formal topology/Structural/Transition gates for final simulation. A dedicated low-resolution simulation mesh is allowed, but it must be mapped to the render mesh and may not be an unexplained Blockout proxy. Colliders need real thickness/contact intent appropriate to the solver.

System Routing

• Thin flexible sheet, clothing, flag, drape: Cloth plus Collision and optional Self Collision.
• Elastic spring-like or volumetric deformation: Soft Body.
• Falling, rolling, stacking, impact: Rigid Body plus passive colliders.
• Hinges, breakable joints, mechanical physics: Rigid Body Constraints.
• Liquid, smoke, fire, inflow, outflow: Fluid domain and flow/effectors.
• Surface response maps and local interaction: Dynamic Paint.
• Deep-water macro waves and foam data: Ocean modifier.
• Local procedural ripples: Wave modifier, Dynamic Paint, or shader detail.
• Static contained liquid with no visible motion: a closed container-conforming volume with physically plausible surface and volume shading; use Fluid only when volumetric motion, splashing, filling, or collision changes the visible result.
• Emission and lifecycle: particle system or Geometry Nodes points.
• Flocking/schooling: Boids when available, Geometry Nodes simulation, curve-guided instances, or a justified hybrid.
• Destruction: installed fracture extension or deterministic fracture geometry plus rigid bodies, constraints, debris, and dust.

Read references/simulation-methods.md before creating a cache.

• For cloth, garments, curtains, flags, sewing, pinning, animated colliders, or fold quality, read references/cloth-production.md before adding the Cloth modifier.
• For contained water, oceans, ripples, pouring liquid, splashes, wet maps, foam, spray, or liquid caches, read references/water-production.md before choosing a system.

Simulation Protocol

1. Probe the current Blender version and installed extensions.
2. Validate units, scale, transforms, topology, thickness, normals, and collider roles.
3. Build a low-resolution isolated test over a short frame range.
4. Record solver, participant, cache, collision, material-response, and field settings.
5. Simulate and inspect penetration, instability, tunneling, explosive bounds, energy gain, and unrealistic synchronization.
6. Adjust one parameter class at a time and repeat up to three technical loops.
7. Bake only after the low-resolution gate passes.
8. Add render subdivision, thickness, secondary debris, spray, foam, or dust after simulation.
9. Preserve the pre-bake setup and report cache paths.
10. Compare the simulated result against the accepted unsimulated silhouette and assembly graph; simulation does not excuse detached mounts, missing seams, or impossible support.

Record each participant's simulation block in construction_graph.json, including system, role, cache policy, low-resolution frame range, penetration metric/threshold, and evidence. A modifier or rigid body without this declaration cannot pass the Systems gate.

Hard Rules

• Do not hand-model gravity folds, impact positions, flow, or fracture when a suitable simulation exists.
• Do not start a final simulation while the source asset still fails topology, transition, support, collision-interface, or wireframe gates.
• Do not weld objects merely because they touch physically.
• Do not assume Cell Fracture or another extension is installed.
• Do not use final-resolution settings for the first test.
• Do not claim realism from a successful bake alone; validate material behavior and motion evidence.
• Do not overwrite an accepted cache without a versioned checkpoint.
• Do not call a plane, open surface, or thin beveled slab a liquid volume.
• Do not infer water from a dark or reflective container interior without reference evidence.
• Do not omit volume absorption when visible liquid thickness materially affects the shot.
• Do not use cloth presets as proof of material identity; calibrate mass, stiffness, damping, collision, topology, and fold scale against the requested fabric and real scene scale.
• Do not add Solidify or render subdivision ahead of Cloth unless the route explicitly requires those evaluated vertices in the simulation and the cost has been tested.
• Do not call a reflective plane a water body, and do not run Mantaflow for a static water volume whose shape does not change.

Completion Gate

Require stable test frames, documented participants, valid cache state, acceptable penetration, bounded velocities and scene bounds, and a review render/contact sheet. Hand off procedural instance systems to blender-procedural-systems and numerical validation to blender-geometry-validation.

For static contained liquid, replace cache requirements with evidence of a closed non-zero volume, container conformity, waterline and contact policy, surface/volume material response, and any visible ripple cause. Preserve the reference hypothesis that justified the liquid.