I wondered alot about why the GR2 editor from TDW did the loose imports the way it does. I spotted the problem after long hours of not thinking of how obj files actually work, but stareing at the exported files from GR2 files.
Then I took a step back and studied how an obj file works, looked at exports from GR2 files, and finally saw at lest part of the GR2 meshes scheme.
To encode a proper diffuse mesh and AO mesh, We need exact count matches for verts and such.
Sounds easy but it's not just that simple.
Well, it is, but it isn't also.
One major thing I found is that a mesh can contain references to a real-face/vert/texture as a placer but it does not get rendered.
It's a kind of OpenGL thing in some ways, but at render time is corrected by the granny engine.
(granny is not the only thing that handles this kind of issue)
We need to open 2 obj files, get a per vertext UV mapping for each.
Then do a bunch of compares and mixing to get 2 matched obj files for import with TDW's Program.
The hard part is what faces grouping can/should be duds to create the best false references?
|