I just tested here 2 embedded quads spaced by 6mm
One have a displacement material of a bit more than 6mm to be able to broke it
So basically I have one face totally plane, and one face displaced.
My goal was to minimise the RAM used by Indigo to have details for the displacement mapping.
But finally the true mesh file from the exemple permit me to go to 20 in subdivision (possible more but I do not need in this exemple), but these dynamic "quads" only 13 and nothing more (with 32Go RAM).
The result is interesting (for me at least ) but it think that we certainly can do better.
- - Someone could explain why this kind of meshes ask more RAM to compute than an igmesh? (igmesh are just a tweaked obj if I am right).
- - My second idea is to build a "air_interface_mesh" from the picture used as displacement mapping (vectorize the bitmap --> extrude the 2D vector to obatin a mesh) and use an "Air" material with 1 as IOR to obtain a nice result.
What do you think about? (okay, this is not a material in such a case, my bad:/)