The volume primitives in the first input to be the hierarchy.

The start of the surfacing operation. This does not have to be on the surface, but it has to be a point that can be settled onto the surface. Specifically, the gradient at this point should point in a direction that hits the surface. If you have a banded SDF, this should be within the band.

The volume value to surface.

Reverses the sense of the iso-offset, effectively inverting the normal of all the polygons generated.

Instead of blending different volumes of the hierarchy, only the volume with the finest resolution of detail will be used. This is best when all the levels represent the same underlying function but they don’t form a proper inclusive hierarchy.

The relative size of the region to blend the interior volumes. This allows for a smooth transition from the lower resolution to higher resolution areas.

Because the voxels of a volume are center-sampled, when evaluating near the boundary of the volume the border conditions of the volumes must be taken into account. This can result in inacurracies, however, especially if another volume could have provided better results. Strict volume bounds shrinks the valid range for volumes to be only those points which do not require any boundary conditions in their computation.

How accurately to settle points to the surface.

When computing the curvature for adaptive step sizing, several gradient operations are performed at nearby points. This controls how nearby the points are. This should be small to avoid skipping features and introducing artificial noise. However, if your scene scale is large and it is too small, numerical error will make everything look artificially flat.

After computing the curvature dependent length, or sampling the edge length from the second input, the desired edge length is then scaled by this number. This provides an overall level of detail to the tesselation.

The edge length will never go below this level. This is very useful when you know what your smallest detail level is.

The edge length will never be longer than this. A maximum edge length is necessary as perfectly flat areas will return infinite edges.

When growing a new triangle, the new triangle will not be shorter than this ratio of the previous edge length, or longer than the reciprocal ratio. For example, with a Max Edge Reduction of 0.5, new triangles will be between 50% and 200% of the previous triangle. This constraint keeps the mesh more uniform and avoids failures in the growing algorithm when lengths change too quickly.

The surfacing operation proceeds by first growing a mesh of triangles and then closing all the resulting gaps. Toggling this disables the second phase, allowing one to inspect the unmodified results of the growing operation. This should always be on.

After the final mesh is completed, an additional pass flips the edges of triangles to try and further improve their regularity.

The point normals are set according to the normalized gradient of the volume hierarchy at each point.

When the surfacing algorithm hits the edges of the largest volume it can either stop there, creating an unclosed surface, or try to continue surfacing. If it is not set to stop, one should ensure the boundary conditions are Constant with a positive value, or are SDF, to avoid the surface from growing forever.