Velvet material

[zukazuka]

I saw something in source control about a new velvet material...

Maybe this thread could be for samples and examples? I haven't tried it myself yet but it sounds cool...

[SATtva]

See here and below for details. Although a dedicated thread is a good thing.

[syndaryl]

Here's one I did last night, baked to just under 1k samples (It's still baking, should have 2k by the time I get home from work I hope). I cleaned up a couple of speckles in GIMP that I'm hoping will bake out.

I also adjusted the colour balance in post here at work, because I didn't have time to adjust lighting colour this morning - I'll be doing that via the light panel of course when I get home and do a final.

Exported via Reality from Studio, manually edited to use velvet and the mapped texture on the tablecloth, instead. Using the Windows binary build from the 18th, windows 7 64 bit.

[SATtva]

Looks like something went terribly wrong with velvet. This is how it renders in 1.0RC1 with directlighting:

Velvet. All default settings except for diffuse color. Objects in the scene are 40 cm high.

The terminator issue is aweful. I wasn't sure about Thickness parameter units. It's 0.1 by default which is extreme if it's supposed to be metric units (or is it unitless?), so i've tried 0.005 as something more reasonable:

Velvet w/ low thickness

Still the same. In any case, it looks nothing close to the original images by tomb when velvet was just introduced. I've also made a couple of reference images:

Matte

Glossy Multibounce

As you can see, the terminator issue is barely visible on the lowpoly sphere, and the highpoly jar looks ok in both cases.

Should i file this as a bug, or maybe i'm missing something here?

[SATtva]

The problem doesn't seem to be related to our recent development -- with Lux 0.8 velvet renders identically. Still, something is really wrong here.

[tomb]

Indeed. I think there is something fundamentally wrong with my implementation of the original paper. Unfortunately, I dont have time to investigate it. Perhaps the velvet brdf should be retired and replaced by multibounce glossy. At least until it is correctly implemented/fixed.

[SATtva]

What i found interesting in your images here is they don't show the terminator issue on torus and cylinder at all. Do you remember whether you've used triangle meshes or shapes?

Perhaps the velvet brdf should be retired and replaced by multibounce glossy. At least until it is correctly implemented/fixed.

Agreed.

[tomb]

What i found interesting in your images here is they don't show the terminator issue on torus and cylinder at all. Do you remember whether you've used triangle meshes or shapes?

Perhaps the velvet brdf should be retired and replaced by multibounce glossy. At least until it is correctly implemented/fixed.

Agreed.

They were triangle meshes. i seldom use shapes during testing. I don't recall what resolution they had though.

Btw. the original paper can be found here if anyone have the time to check my implementation: https://docs.google.com/viewer?a=v&q=ca ... jIeWmwhkOA