ptex texture sizing and subdivision

[Zach Gray]

I'm having trouble with texture sizing with subdivided ptex files.

Is this issue with how mudbox sizes the textures, or with how vray renders and subdivs?

1,2 show WDFA mapping solution


3, shows my result in vRay with 3 levels of subdivision, you can clearly see the stretch.

4, shows my painted object, in mudbox, with 3 levels of subdivision

5. shows base mesh, with/without subdivs.

[Metzger]

I have the same issues when sub-dividing ptex from Mari. I believe this to be a V-Ray issue. One work around would be to model in more edges.

I hope they can get a fix for the stable build.

[Zach Gray]

For now, I'm just smoothing the geometry before assigning ptex.

When I look at the models in the Disney PDF, they are really nice and sparse.

http://www.disneyanimation.com/libra...tex-slides.pdf

[Zach Gray]

The more I review the information on Disney's site, the more I think the entire picture may be incomplete by chaos and mudbox/mari.

It looks as if there is a custom anisotropic filter in prMan, and from my testing, the texture distribution workflow in mudbox is pretty incomplete compared to WDFA's paint 3d. Didn't the foundry license some of paint 3D?



Getting the file format implemented is a good first step. Hopefully the vendors can eventually refine and add multi channel support.

[Metzger]

I heard the Foundry and Disney have something going on, but I have no idea to what extent that is. I do know that there's something fishy when it comes to rendering sub-div ptex in V-Ray.

There's multi channel support in V-Ray, but none in exporting with Mari. Also currently I think Ptex only supports channel IDs. So you would have channels with numeric descriptors only. Couldn't name a channel like spec.r for instance as you could with exr channels. I could be wrong but that's what I was told.

-Scott

[Zach Gray]

I heard back from the mudbox developers. Subdivision should be properly accounted for on that end, but I need to double check my workflow, I think I set up ptex before creating the subdiv. I'll post as soon as I test.

When distributing texels, Mudbox takes into account the size of faces *at the highest level*. So for a subd workflow be sure to subdivide the model several times (and either sculpt or apply any displacements) *before* setting up ptex.

[Zach Gray]

My workflow was incorrect in mudbox. You definitely need to smooth before setting up ptex.

But the filtering scheme in vray isn't working correctly, and doesn't seem to be following the anisotropic approach in the disney presentation of the prman shader. (I'm probably wrong about this, just observing what I see.)

http://www.disneyanimation.com/library/ptex/ptex-slides.pdf


Box filter 1.0 with mip levels filtering. 4 subdivs


No filtering. 4 subdivs


No subdivs, showing baked in stretching from mudbox - correct behavior.


Mudbox ref image, cube with 4 levels of subd.

[VasilVelikov]

Doesn't look nice. Would you mind attaching the scene with the cube and the .ptx file so I can test it and fix it eventually?

[Zach Gray]

....scene with the cube and the .ptx file so I can test it and fix it eventually?

http://southernlabs.net/downloads/vray/


I uploaded a monster ptex, 350mb with a high contrast cloth pattern so you can check for aliasing artifacts.

The cube and map should have 5 subdiv levels.

[Zach Gray]

Doesn't look nice. Would you mind attaching the scene with the cube and the .ptx file so I can test it and fix it eventually?

Any updates on this? We're under the EDU builds, so I don't have access to the changelogs via nightly builds, so I can't look for the update on my own, and Vlado has to do a special build.

I'm moving forward with projects using ptex in good faith that this will be solved sometime soon.

Thanks! I appreciate your innovation and early adoption of new approaches.

[vlado]

It is not quite fixed yet, Vassil is still working on it; from what we discussed it should not be too hard to fix.

Best regards,
Vlado

[VasilVelikov]

I have fixed the filtering a couple of days ago and it should work fine now. If you notice any more problems please report them.

[Zach Gray]

Thank you!

Can you enable 'vray allow negative colors' on the ptex node? Using ptex isn't useful for displacement from mudbox without that.

Zach

[Zach Gray]

Vasil,

I've done a bit more testing on the negative colors issue. I think the vrayPtex node clamps the output 0-1 when displacing, and the mudbox extract maps has negative colors and 1+ colors.

I tried hacking the attr in there, but it didn't affect displacement.

 vrayAddAttr "ptex1" vrayFileAllowNegColors;

When I hooked the map to the diffuse channel, the positive displacement was white, while the negative was black. I fed it through a +/-/average node, and could see the different values. That seems a bit strange, because I could get to the values this way, but not when the ptex was hooked to displacement.

[Zach Gray]

Mudbox


Vray, ptex feeding into diffuse via set range to preview map. It seems correct.



Vray, ptex -> disp. Clipping 0-1 (I assume the other artifacts are from improper filtering, I don't have update yet)

neg_map.ptx 30MB
neg_object.obj 12K


http://southernlabs.net/downloads/vray/