Streamline backplate integration workflow

[^Lele^]

I posted you both versions of the file, and instructions to get the results you want.
I showed you, further, V-Ray do precisely what you wanted: never it crossed my mind that you could live with a throwaway alpha channel in your renders.
Therein lied my issue: with tying the two together: the very first render i posted had the plane not receiving GI, hence not darkening, and the alpha correct for compositing.
here's a link, to jog your memory:
http://forums.chaosgroup.com/showthr...547#post675547
here's the two images:


You said it was wrong, because it didn't receive GI.
Here:
http://forums.chaosgroup.com/showthr...557#post675557

Problem 2: You have masked the GI darkening bug by disabling matte objects to be receiving GI, and that's exactly the problem, because if you then integrate some glowing objects into your scene, such as glowing sphere, they won't illuminate your backplate, they will just have dark black shadow.

That is what started my confusion, and the chase of the wild goose you sent me on.
Every other renderer works precisely the same, and there are images to prove it, for those too lazy to try themselves with one sphere and one plane in half a scientific and objective manner.
We can keep on debating on who's wrong, and who's right, but that will not avail your needs, certainly not more than i strived to that end thus far.

I shall pass the issue on to those better than I at this, and ungracefully bow out of it.

[LudvikKoutny]

Oh and just for fun, even though as I said, alpha doesn't matter if you dont render it on black don't plan to comp it in post anyway, if you want to get right alpha in the fixed scene i sent you back, just drop Environment/Background camera map node between Matte material and HDRI map



Just to prove it's possible

[LudvikKoutny]

Ok...

I can't believe I have so much patience, but here we go again. Please this time, try to read my post thoroughly:

The scene you have sent me which you claim is working does not really work, because it has disabled receiving of GI:


If I render that scene with Mental Ray, I can get it all working, with GI:


If I enable receiving of GI in your original scene, I get that wrong GI darkening:



Now please be so kind, and do read what I already mentioned in my previous post:
There seems to be another bug in the V-Rays matte solution, that causes matte object to receive a lot more GI from self illuminated objects, than it should. The blue teapot in the scene I posted has intensity of only one image unit, yet it emits significantly more light. So the global illumination in Corona and Mental Ray version of the scene is present, just a bit dimmer.

Here is the same scene, I have just increased Self-Illumination intensity 10 times:

Corona:


Mental Ray:


Obviously something is wrong with V-Ray. If you have illuminator under direct sunlight and it emits just enough energy to be dark blue color, then it can not emit brighter energy around itself than the brightness it itself is. That would break energy conservation.

[LudvikKoutny]

And here's additional proof it's just another bug.

This is what matte ground plane looks like:


Now this is what it looks like when i disable mate, and just project the map on a regular diffuse material of very similar albedo:


Notice how suddenly teapot of same self illuminated intensity emits a LOT less light... About as much as Corona and Mental Ray reference renders, huh?

EDIT: I am a little sad you edited out the part of your post about reference renders not having mattes catching GI. Now my hunt for the self-illum bug seems out of context :/

[squintnic]

have you tried to use the matte properties

[Dave_Wortley]

Lele, in the same way there was a simplified render settings dialog made for people with less than expert knowledge to use there should be a dialog for loading backplate and making ground plane with environment maps set up correctly.

I'm not sure we've ever gotten rid of that darkening problem either, every time I go to set it up I always play around trying to get it better and I've chatted to Vlado about the complexity factor of it many times before.

[kosso_olli]

The video posted by Recon422 could not be better, it very well illustrates the problems I have with Vray and car renders every day. It is way to much clicking in my eyes, and it has to be made simpler and correct. Lele, get into this discussion again, because there obviously is something very wrong. As I said before, tell us what we need to click in order to get results like Corona.

Edit: Recon422 has posted a scene that clearly illustrates the problem. How hard can it be to profile that and trace the bugs that are in there?

[^Lele^]

BackplateSetup_final.zip

The file attached is a setup that ticks all the boxes Recon wanted ticked, AND it keeps the alpha fully usable.

proof:


IT involves setting up the mattes normally (either through the vrayProperties or a MtlWrapper), like so:


The backplate goes into the environment override for the secondary mattes.
And the trick, the solution to the conundrum, was dangling right in front of my eyes all this time, and yet it's Vlado that ultimately told me how to get it right:
One tickbox to enable the Environment GI override, and then one sets it to black.

like so:


and this is it.

This will be changed, in code, in the next builds, as this setup in fact allows the full latitude of cases to be covered (ie. all the light interactions AND the compositable alpha channel).

[LudvikKoutny]

You are probably going to kill me, but I am still not happy:

The solution of enabling GI override and setting it to black was already brought up on page 3, in post #30: http://forums.chaosgroup.com/showthr...061#post674061

While this is the best "Workaround" so far, it comes with issue of its own:

As soon as you enable GI override, splotches will start to appear in contact areas between CG objects and plates:


Those splotches are especially obvious when surface of the backplate does not have much detail, such as constant or nearly constant color - which can happen often in real world:


It even seems to prevent the projection for GI issue, but again, at the cost of very obvious splotchy artifacts:


And I can also see enabling that Override still does not take care of the issue of self illuminated objects spilling a lot more light on mattes than energy conservation would allow for. Will that be fixed too?

[^Lele^]

The GI splotches are an unfortunate side effect of the Leak Prevention Algorithm in the LightCache.
While it may or may not be fixed, the splotches disappear if you either turn leak prevention off (0.0 in the spinner) or if you choose BF as secondary (if you're doing exteriors, there'll hardly be any speed difference.).
Below the effect with it on and off, in the alpha is particularly visible, regardless of plate.


As to

"the issue of self illuminated objects spilling a lot more light on mattes than energy conservation would allow for"

i don't see that, and there would be absolutely no theoretical reason for them not to behave correctly, under the circumstances.
Turning the GI ENVIRONMENT off does absolutely nothing to the GI generated from looking up the dome light (but doesn't do the job twice, avoiding the darkening and resulting in exactly the same lighting as if the option was on.), nor it does a thing to the GI generated by the objects' albedos, be those below or above 1.0.
In fact, here's proof of what i am saying: the green emitting sphere makes a larger pool when the plane is shaded as non-matte, white diffuse.
With the backplate projected, the pool is smaller, so the talk of energy preservation being broken is utter nonsense.
Rather, it follows the irradiance*albedo rule quite well, given the relative brightnesses of the stuff that's being lit (you could call the white diffuse render a NORMAL render, in the scientifical sense, for the light distribution on the plane, as the light's irradiance is multiplied by the 1.0 of the surface albedo).

[LudvikKoutny]

Alright. Turning off leak prevention definitely did the trick. Although my original request of streamlining the workflow still stands. Now we have two additional steps. Enabling GI override, and changing value of a feature that's buried in the expert mode of UI.

Here's a scene showcasing just the self illumination bug:

Two teapots sharing same self illuminated material on top of two planes of exactly the same albedo. Plane on the right is matte object, plane on the left is just regular diffuse material with disabled matte:


Scene is here (it uses the same morning backyard HDRI we've been using so far):
MatteSelfIllumBug.zip

EDIT: Come on man... You yourself accused me of my testing not being scientific, yet here you post an example going "Oh look, if I disable the matte and get a ground plane of a LOT higher albedo, light spill will be a lot higher too!" Well of couse it will! Because that bright gray color has significantly higher albedo than the fence background! You need to compensate for that by adjusting the color so it at least roughly matches color of the fence behind.

[^Lele^]

read my post above.
That is the correct behavior.
in fact, make the material in your plane WHITE, and not a murky 0.25 gray, and the light from the teapot will be bigger than the one on the matte object, as correspondingly the shaded color is brighter.
like so:

The ultimate proof comes from having everything for repros as white, and the diffuse color as white (and then expose down to see the lighting):

[LudvikKoutny]

Even though I would dare to call my top view scene a bit more controlled environment than yours, even in your scene, it's easily replicable as soon as you (at least roughly) match the diffuse albedo of ground plane to the color of the fence behind:

Original:


Diffuse:

[^Lele^]

dude, are you serious, now?
Take at least the time to read the replies, as a sign of decent respect for the time you're taking up.

[LudvikKoutny]

Well, you tell me?

When was the last time you were integrating glowing objects onto 100% white backplate? Really?

The simple point is that if I do CG integration onto a backplate, and i put self illuminated object of multiplier 10 on the plate, I expect to get illumination adequate to the albedo of the object, like in case of Corona, or Mental Ray:

Corona:


Mental Ray:


Not a nuclear explosion like in case of V-Ray:


What's the point of going personal with phrases like "dude, are you serious, now?" I think it should be in your best interest to get the the bottom of this as it is mine. If I dare to speak for other users, I think they I would appreciate if self illumination on mattes simply worked predictably without nasty surprises.

Yes indeed, if you open my scene, and make left plane a lot brighter, so it's albedo is not matching one of the plane on the right, then indeed you will get larger light spill. That's how physics work.

At this point, I am done. If you work for ChaosGroup and your opinion is that there is no issue and you are going as personal with it, then I can't do otherwise than suck it up. I am sure everyone else will be "happy" about that as well.