3D Geometry

#16 ChadCapeland

Flying Pig

Group: Adv Members
Posts: 834
Joined: 15-April 04

Posted 03 August 2010 - 09:57 AM

Yeah, we looked into doing PC2 and BIN files. That's how we know about the normals thing. The problem is the FBX Loader, we could replace it entirely with our own tool, but we figured we'd wait for Eyeon to fix it.

Regarding alembic, Rob Bredow indicated that Maya would likely have native support for it with the next version, and it seems reasonable to assume that other ADSK products would follow, and Houdini has a good history with supporting stuff like this, including Field3D. So this seems like it might be a MUCH better solution than the blackbox FBX format. I would like to see Eyeon keep an eye on this.

- Chad

#17 Daniel Koch

Flying Pig

Group: Adv Members
Posts: 627
Joined: 21-March 04

Posted 03 August 2010 - 10:00 PM

Thing about MDD is it's apparently missing more than just normals, but even faces too - it's literally just a collection of times and points, if the spec I saw is accurate. You could use it for motion paths or point clouds, but nothing else.

Alembic looks a lot more capable. We'd be a little hesitant to adopt another 3D format (FBX support already takes a sizeable chunk of our resources), but we're looking at it.

#18 robocop

Flying Pig

Group: Adv Members
Posts: 248
Joined: 27-February 09

Posted 04 August 2010 - 03:53 AM

Quote

We'd be a little hesitant to adopt another 3D format

May the Force be with Eyeon!

#19 protean

Flying Pig

Group: Adv Members
Posts: 181
Joined: 15-July 08

Posted 04 August 2010 - 04:23 AM

Daniel Koch, on 03 August 2010 - 10:00 PM, said:

I think we might have our wires crossed. MDD is a deformation 'parameter' for want of a better definition. It is not an object description like OBJ or FBX. The idea is that it's to be used in a deformation plugin on an imported geometry.

I know Fusion expects an object sequence (everything defined every frame) but what I'm suggesting is importing an object in rest pose and applying MDD as a deformer per frame.

J

#20 ChadCapeland

Flying Pig

Group: Adv Members
Posts: 834
Joined: 15-April 04

Posted 04 August 2010 - 08:45 AM

Daniel Koch, on 03 August 2010 - 10:00 PM, said:

Which is fine, provided your FBX Loader kept the vertex indices the same. Storing only an array of positions is quite compact, which is the whole point. Unfortunately, the FBX Loader does not preserve topology, so you can't get these to work together (without perhaps rebuilding the mesh after the FBX Loader with some sort of calibration MDD).

Daniel Koch, on 03 August 2010 - 10:00 PM, said:

Alembic looks a lot more capable. We'd be a little hesitant to adopt another 3D format (FBX support already takes a sizeable chunk of our resources), but we're looking at it.

I only mention because A) Maya, Houdini, Nuke, and assumedly 3ds max and XSI will all support it, which means it could become important to customers, and B) it's not a black box like FBX is, where you are hoping that it works, and when it doesn't you are sorta stuck. ADSK can release a new FBX tomorrow which works in 3ds max, Maya, and XSI perfectly fine, and you'd have to figure out how to make it work in Fusion really quickly but without anything to go on other than the libraries. I suppose you could add C) it actually does what a compositor wants it to do, which is to pull point/mesh/surface/camera data in without caring about the underlying structure. FBX wants you to get a mesh and skinning weights and bones, which as far as a compositor is concerned is stupid. An independent surface description is sooooo much more useful, in the same way that the average compositor wants to get EXRs, not RIBs.

- Chad