noob questions about deep

[LeoYfver]

1. In arnold have some more deepexr options like tolerance. Do we have the same type of control in vray?





2. We have some choices on the RE settings, are these just for the aovs and not the deep and RGB? not sure exactly how these works and on what. For example if i dont have any elements but only deep and RGB how do i control it?

[LeoYfver]

So one issue is when i try to deep merge and i get an aliased edge. Is this just how deep works or is it something i can do in vray to get better result in nuke.

Left is the sphere with grid as holdout and the grid with the sphere as holdout deep merged. right is a beauty with both of them together.

[LeoYfver]

I will add another question here :P

we want to render out the beauty as default and the a separate deep channel. Now it seems like it render the RGBA+deep instead of only the deep.
I tried to tick dont save RGB channel but nothing happened.

[david.anastacio]

Hi Leo,

We have bumped into the same issue, after a long time getting complains from the comp department about how many few deep samples the renders had, one of my fellow CG Sups here finally had some time to take a deep dive into the issue and finally make some progress on it.

By default, just turning on the deep in the ROP will not give you good enough sampling, it needs some AOVs to work correctly, either a render id or a zdepth AOV depending on the deep settings in the AOV channel container.
See: https://docs.chaosgroup.com/display/...nels+Container

We got very good results by setting "Deep Merge Mode" in the "V-Ray Channels Container" to "By Z-Depth" and lowering the value of the "Deep Merge Z-depth Threshold" to 0.1 (go even lower for more samples, but this value was woking good with motion blur even), and also connecting a ZDepth AOV to that Channel Container.

We tried the merge by render ID and using a render ID AOV node and that didn't seem to work correctly, bug most likely. We had a chat with Gosho today about these Deep "issues" , so the devs are informed.

I think the big issue here was the lack of information, this workflow was not very clear and having the deep settings broken into to places (ROP and Channel container) does not help, hopefully this will be improved in the future.
Also, and very important for anyone else that might know this yet, deep does not work when using cryptomattes.


Example.

V-Ray render from Houdini with depth of field and motion blur:


Channel Container setup in Houdini:


Intersected gray checkerboard plane in Nuke using a deep merge:


200% zoom of the intersection:


Deep samples read out of the motion blur area (has lots of samples):


Best,
David.

[LeoYfver]

david.anastacio thanks alot for the explanation and it helps alot. We come to the conclusion to use almost exactly the same settings but the added zdepth aov was neat to know.

about cryptomatte it makes sense it wont work but would have been nice to get the same id number and some metadata as a kind of deep cryptomatte. But i dont know if that would work either :P

Regards Alexis