As far as I can tell, there are two big issues with converting Heightmaps to geometry in Houdini. I was wondering whether there are any new solution for these.

The way I do it now: I load a picture into cops (and manipulate it). I create a grid, uv it and assign height via a pointVOP. Next I polyExtrude the geometry and convert to volume.

1.: I create a grid with the same amount of points as pixles in my map. So at 2048*2048 points Houdini starts getting slow. At 4k*4k it is almost impossible to work. Is there a more efficient way to transfer a height map into geometry in Houdini (while maintaining detail)?

2.: For large maps I would treat pieces of them separately. This would allow me to create smaller more manageable pieces of geometry. The question is how to make the chunks align perfectly at the edges. During conversion to and from volumes (and subsequent poly reduction), inconsistencies are introduced along the edge of the meshes. How could this be avoided in the first place? How could it be fixed efficiently?

- My idea here is to code some VDB magic to import the heightmap directly into VDB without the in-between via grid and vops and extrusion. This should save tons of performance and reduce the amount of inconsistencies gained. - But I don't really know how to do it.

There's a method presented by Ari Danesh in the Go Procedural Terrain Generation tutorials on Vimeo:

Houdini Projects: Terrain Generation 1: https://vimeo.com/116364169 [vimeo.com]

about 1hr 3m in, he uses 3 grids and a switch; a low res for low poly game export, base res for viewport display, and high res for rendering out.

Thanks! The tutorial gave me some good ideas. I had a specific purpose in mind for using vdb, but now I am thinking I should work that out in a different way.

- I am still interested in a method of going directly from cops to vdb (without going via vops and grid), though.