How hard would it be to add a VDB writer? This could be used with Houdini.
Here is the e-mail request from my user:
VDB is an export standard which Houdini can read in.
Export of .geo format data only works for polygons. What we want to do is export volumetric data in VDB format so that we can perform volumetric post-processing in Houdini.
There is a package called Redshift which allows you to perform volumetric rendering of VDB data in Houdini using the GPU. Render times per frame have been shown to go down from 1 hour to 2 minutes.
From my perspective, this is a pretty big WOW.
So, my questions:
- Does it make sense to add this writer into ParaView?
- How hard would it be?
- Can we get it in for the 5.7 release?
It’s doable for sure. We’d need a bit more exhaustive investigation to see how complicated it could be, but shouldn’t be too hard esp. if we can start with a small set of use-cases to support in the first pass.
Sounds good. I will write it up.
VDB is now also implemented in cinema 4D. It would be amazing to have this voxel format in paraview!
OpebVDB is open source, C++. Every 3d package that I can think of supports it. Maya, blender, modo, lightwave, 3d studio max, cinema 4d, Houdini… Also render engines like renderman, Arnold, vray and others support it.
I assume that with all of that support, it is not too difficult to create a writer.
The writeup is here: https://gitlab.kitware.com/paraview/paraview/issues/18501. I am changing priority for the 5.8 relase (i.e., Fall) to required.
I see that the VDB Writer is already in the works here. That said, is there a VDB Reader for ParaView that is available somewhere? Or, would it be part of the Writer addition?
Afaik, these is no .vdb reader in ParaView
Yes, please, make both writer and reader. It would be great combining it with Blender!
Very preliminary work is at https://gitlab.kitware.com/paraview/paraview/-/merge_requests/4855 for those that want to check out an OpenVDB writer in PV. Outputting from vtkImageData has made the most progress so far.
Is a reader planned as well ?
I’m not going to develop a reader. Since VDB is designed to be a sparse storage format with multiple “grids” it’s not clear to me what type of dataset to read in as. The best I can think of right now would be a composite dataset of point clouds. We don’t have a strong use case for reading in VDB files into VTK or ParaView so we can’t justify spending the resources to develop a reader.
Ok, I think that will be the first writer without a corresponding reader.