Dan Wood

Can't check over it myself at the moment, but I'm wondering if the group-expression version of the intersect command might be the way to go.


If you have each mesh assigned to a primitive group beforehand, you may be able to use a standard group pattern in that (string group) parameter... along the lines of "* ^thismesh" for intersecting everything excluding the current mesh.

Assuming you want to cull on a per-connected-mesh basis, rather than a per-point/primitive basis... you could perhaps start with the intersect vex command, make a check along the normal from each point in a mesh, store whether it hits anything or not... then use the promote note to aggregate the hits per mesh, convert number of hits into a percentage of total point count, and set a threshold to delete the whole mesh, say if 90% of intersect checks hit something, then delete that mesh.

I suppose the point is, if you went to accurate lengths to replicate the simplicity of a simple vfx scene in reality, it would look like a simple vfx scene - If you reproduced this geometry precisely, with no additional smoothing/rounding, from a consistent, entirely opaque, untextured, non-reflective material, and pointed a single light-source at it in an otherwise dark room, then yes, in real life the shading would be like this.

Reality doesn't look like that because reality tends to involve fabrication processes that yield smooth radius-of-curvature transitions (eg, using sandpaper), complex materials with internal light scattering properties, and complex lighting environments with various primary and secondary light contributions coming from all around.

If you seek to model all of the subtle complexities, you'll get a realistic looking result.