Zero out the density and temperature that are inside the collision SDF, avoiding any leaking of smoke.

The output velocity VDB v will be scaled by timescale, ensuring it will motion blur correctly. However, if this is to be used by another simulation, you may want the raw velocity values.

The output velocity VDB v will be adjusted by ambient velocity, ensuring it will motion blur correctly. However, if this is to be used by another simulation, you may want the raw velocity values.

A global ambient velocity that the simulation operates within. All the fields will advect with this velocity, allowing it to provide a global wind effect that can be directly keyframed. However, the actual velocity VDB doesn’t contain this.

The amount to scale the time inside the block. A value of 1 maintains the normal speed, a value greater than 1 speeds up the simulation, and a value less than 1 slows down the simulation.

Each frame is broken into a series of substeps. This can give more accurate results for fast moving simulations, but at the cost of increased time.

Turns on live simulation to provide an interactive, non-deterministic simulation mode. It simulates independent of the playbar and isn’t associated with any keyframes or caching.

This parameter (and its associated parameters) is intended as a sandbox mode where you don’t have to rely on the playbar.

When Simulate is on, this is the amount to scale the time inside the block. A value of 1 maintains the normal speed, a value greater than 1 speeds up the simulation, and a value less than 1 slows down the simulation.

When Live Simulation is on, this lets you pause and resume the live simulation.

When Live Simulation is on, this is the current tick (frame) within the live simulation.

The projection process is repeated this number of times. Increased iterations takes longer, but will produce a more accurate result. It will also increase the faithfulness of the collision operation.

Size of the kernel to apply at the voxel size. Larger will be higher quality but at a rapidly increasing cost in time.

Apply two voxel kernel passes before performing a block pass.

Enable the block pass. In this pass, divergence is reduced to dipoles in large blocks, allowing action at distance without having to increase the voxel kernel size.

The size, in blocks, to accumulate the surrounding block’s dipoles when updating a voxel. Larger will be higher quality but at a rapidly increasing cost in time.

Reduce the strength of the block correction over the farthest blocks. This avoids sharp transitions in the corrected velocity if there are strong impulses, but will require more iterations or time to correct divergence fully.

Combine the block dipoles into an even coarser set of points. These are then applied to every voxel. This ensures global effect of divergence in a single pass.

Projecting using corner-sampled divergences can miss certain modes of the velocity. The hourglass filter applies a smoothing operation to ensure these are correctly filtered; this can reduce jaggy artifacts.