Multimatte consider for anti-aliasing not working

[rainingwaffles]

When rendering a multimatte it seems to often get grainy samples in areas where two multimatte channels overlap, even using the "consider for anti-aliasing" option.

The two attached images show an example where the multimatte becomes very grainy in areas where two motion blurred objects overlap (the two planes are *not* on top of each other, by the way). The simple solution is to assign different shaders to the objects in the RGB, but this is only a simple example of a problem we see in more complicated renders where the solution isn't always as simple. A zip file is attached with vrscene and maya scene file to recreate the example.

This was tested in V-Ray 20713... has it already been fixed in later builds? Thanks!

Blake

[vlado]

No, it's not fixed. There is a technical issue there that I'm not sure how to solve yet. For the moment, V-Ray considers the intensity of the render elements to decide if it should sample more; however the different components of a multimatte element all have the same intensity, since they are pure colors in one component. So to V-Ray they all look the same.

It will work better if you use the "Adaptive subdivision" image sampler, but that may affect render times too.

Best regards,
Vlado

[mr furious]

i've noticed that the multimatte AA sampling is tied to what is being sampled in beauty RGB of the image. so if there is info in the RGB then i get good sampling in the multi mattes. (not that i am arguing with vlado!

the examples above with solid colors there is not difference RGB wise between the two squares where they over lap. but the edges are fine.

what i tend to do for multi mattes like this is make a render layer that is a copy of my beauty layer (with all bty shaders) and remove all the lights. Then go into the vray globals and turn on default light for that layer, turn off shadows, turn off raytracing...etc.. so vray will calculate the beauty RGB really quickly and apply that filtering to the multi mattes...i no longer have the same issues you have.

its a work around but it works.

my .02

Andy

[rainingwaffles]

Ahh, very interesting, Vlado. I never realized that intensity was the sole factor for anti-aliasing in MultiMattes, and it looks like that's the same criteria for ExtraTex (image attached). Not only that, it appears to only consider anti-alias once based on the combination of *all* ExtraTex and once for the total of *all* MultiMattes, so splitting channels into separate ExtraTex or MultiMattes results in the same noise on the overlapping areas.

In the end, the ability to anti-alias based on MultiMatte and ExtraTex at all is still a big benefit and this limitation isn't a huge deal. Knowing the criteria will make it easier to tune around. Thanks for the explanation...

[Morbid Angel]

what we do is we render all utility passes separately, so we get customized anti-aliasing/sampling for them. Otherwise you just need to increase the min sampling and it should solve this problem but of course increase render time.