Streamline backplate integration workflow

[piotrus3333]

[LudvikKoutny]

Ok, this finally works, but how is that making any sense??

It also comes with a few unexplained issues:

1, How can GI have such impact on scene when entire scene is enclosed inside of DomeLight? GI should not be able to get through domelight?

2, How can GI ever darken the scene? If the backplate is solid white, and domelight is solid white, how can Global Illumination actually subtract energy, when the essence of Global Illumination is strictly additive light transport?

3, Why, when I activate the override, Light Cache starts to make blotches in contact areas, even when retracing is enabled?:


4, This adds yet another obscure step to already quite extensive amount of steps required to set up such basic backplate integration, so original request of streamlining of backplate integration workflow still stands.

Now we finally know there's one last random button we need to enable for workflow to start (somewhat) working, but that doesn't change anything about the weirdness of the problem. It's just partially functional workaround.

[piotrus3333]

This option should be enabled all the time if you are using dome lights and put backplates on renders. It is in fact sort of a bug in dome - dome still passes through lighting from max's environement. You just don't see it most of the time because of vray cam exposure. I remember a version of vray manual saying that dome blocks other environements but that is not true. I just accepted the fact.

why is it darkening mattes? no clue, probably bug, more obvious this time.

[LudvikKoutny]

It is in fact sort of a bug in dome - dome still passes through lighting from max's environement. You just don't see it most of the time because of vray cam exposure.

This should not be true, at least I really do hope it is not, because otherwise I could be messing up all the time, as I have same map that's in DomeLight duplicated also in environment slot, so that I see environment map as viewport background and can rotate it with realtime feedback.

As long as V-Ray DomeLight is set to full sphere, there should be no way for GI to get through. You can test that very easily: Create a white, full sphere dome light and set Max's environment color to 100% red (RGB 255,0,0) then render the image and run color probe over it. My image did not have even slightest red hue, so GI is not leaking through. But with matte objects, something very weird starts to happen indeed.

I really do hope that matte objects are only case when DomeLight does not completely override Max's native environment.

Would be good to have some official response on the matter

[piotrus3333]

You're right. Environement is getting through half dome only. I don't use spherical dome lights (arch viz).

[jasonhuang1115]

Curious to learn if anyone from the dev team can confirm if this is a known issue or bug that will be addressed. And if this is Max specific or this issue is also present in Maya?
Any workaround maybe?

[LudvikKoutny]

Bump...

Now that 3.3 is out, pre-release crunch should be over, and this is a very serious issue that really needs to be addressed. I don't want to be switching to some other renderer every time a job with some backplate/footage integration arrives. Since V-Ray now supposedly focuses on niche market, as essential VFX feature as backplate/footage integration toolset should work flawlessly... or at very least, should work.

[Morne]

I've got a HUGE upcoming job where I need a LOT of this also. Would be good if this issue can be fixed

[kosso_olli]

Still waiting for a reply from the Chaos staff, too.

[^Lele^]

MIS is sort of a hack to get faster/cleaner lighting.

Multiple Importance Sampling is actually the foundation of every single raytracer currently on the market.
https://graphics.stanford.edu/course...h-chapter9.pdf
Turning it off can have devastating consequences on rendertimes, and likely you haven't seen those due to not reaching a critical point with geometry and light counts (and in no small part to V-Ray being quite resilient and quick as a baseline).
In less optimised solutions (say, renderman 16) this would be able to make renders go hundreds of times faster (or slower...), when the subject was miles worth of hyper-detailed geometry (without many of the usual rMan reyes tricks at the avail to help.).

Leave the MIS checkbox alone, for your own sake!

Back on topic:
I read through the thread, but am a tad confused.
Could someone sum it up for me in one post, with a scene to play with, as if they were talking to a fairly slow two years old?

1) What is the ultimate, overarching goal of the exercise? Insertion of CG on a backplate, with shadows and GI from an HDRI visible on the ground plane?

Please be as succinct and precise as you can be in explaining this, for of the half-a-million ways to skin the same cat, i need to understand exactly which one we're talking about, here.

I may have other questions later, but i would need this answered unequivocally (for me!) first.

thanks for the kind patience with a slow donkey! ^^

[LudvikKoutny]

Ultimate goal is following:

Insertion of CG elements onto a backplate with matching HDRI. It needs to be able to handle high contrast HDRI's with for example sun, so dome light is crucial (no fake manual placement of additional lights). Backplate should be correctly projected on the ground geometry so that both reflections and contact color bleeding are correct at any place. Global Illumination is necessary in case integrated elements had some glowing components that will illuminate the backplate/footage we are integrating or CG objects into.

Last but not least, other part of the goal is that this solution is easy, with as few steps as possible. Now this is very important: There should not be any obscure steps involved, so that when you have HDRI with let's say 20 backplates in a set, creating a new integration for different backplate should need just backplate replacement, matching of new camera on it, and rotation of HDRI to match the backplate angle. Only these three steps. Any additional step is an overhead resulting in a larger room for error.

I am not looking for any workaround here, I am looking for bugs to get fixed. In this case, there is only a single way to skin that Cat, and it should be as easy, fast and problem free as possible.

[^Lele^]

thank you. (for now. )

[LudvikKoutny]

By the way, I've posted example scenes few pages back. I can post them again if you want. Actually, current workflow V-Ray has for matte/shadow integration is just fine. There's just pesky bug that GI darkens image when using dome light (but not without it), and that GI leaks through objects when backplate is not used through Camera Map Per Pixel map (which adds lot of annoying steps into the workflow). So once these two bugs are squished, it's gonna work perfectly, and everyone will be happy. No need to re-invent the workflow.

[^Lele^]

Let me know if this solves your issues.

I lacked a matching plate for the dome, so i only used the dome.
I'm confident you'll substitute the plate in the right places without issues.

The red is coming from the vertical boxes, so to show it's correctly calculated (at least in emission on your subjects.).
Feel free to replace the shader with the one on the ground plane, so that it will project out the colors of the camera mapped HDRI (this is an actual technique to build HDRI-projected geo and give VRay the ability to light scenes with also positional infos.).
Another thing to notice: i am bumping the shadows with the cam-projected HDRI as bumpmap.
So if your plate had a ground made of shiny tarmak, with plenty of fine detail, you'd also get that much better of an interaction from it.

Once we established that there is no bug to fix, or which is the bug to be fixed, we can start looking at automating at least some steps.
But saving out the main override shader, and swapping out plates is all you'd need (well, on top of the geo properties, ofc.).

If you want the GI to actually LIGHT your reprojected plate, well, then it will be unavoidable for it to be visible in the calculations, for light will fall on it, and correctly so, provided the geometry matched the underlying HDRI well enough (again, this is a technique to make CG interact with the plate.).
Of course, rendering with the correct (additively so!) render elements, mtlSelect, and multi-mattes, will allow you to split the components just as you want, so that in post you'll be able to drive with precision and correctness the various contributions.

Let me know of your thoughts, please.



BackplateSetup.zip

[LudvikKoutny]

Hi,

unfortunately you have completely misunderstood point of the thread.

First of all, material on your ground plane contains some crazy OverrideMTL hack, and you do use Camera Map Per Pixel map. Camera Map Per Pixel map usage is problematic as every time you switch camera in you already built scene, you have to manually pick the camera for backplate depending on viewing angle. So there goes simplicity. Imagine you have your scene already built with 10 cameras, and you switch them around and always have this annoying step of picking camera for the Camera Map Per Pixel map.

So that is the problem 1, which I already described on previous pages: You MUST use Camera Map Per Pixel map for Backplate to map properly for all ray types. Just enabling matte object will automatically project backplate for reflection/refraction rays, but not for GI rays. Resolving of this but would consist of correctly projecting backplate on top of matte geometry for all ray types.

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:



If you enable receiving of GI on the matte objects (in V-Ray properties), so that objects correctly illuminate the backplate, then you will get that weird GI+DomeLight darkening bug:



Darkening Bug is caused by using GI in conjunction with DomeLight. If you disable DomeLight, and let GI only sample the environment, your GI will not darken matte objects anymore, but GI on it's own is not able to handle high contrast HDRIs so you will lose hard shadows:


Yes, if you want LIGHT to fall onto the backplate, then yes, it should be visible. But GI should not DARKEN the backplate. GI, by it's nature, can not ever darken anything. It is strictly additive light transport. Yet in V-Ray, if you use GI along with DomeLight to light matte object, GI DARKENS the backplate, and even in cases, where there are no objects on the backplate to occlude backplate from environment:


If you just enable receiving of GI for your matte objects in your scene, and render just empty ground plane, you will see the darkening, even though there is nothing that could possibly darken the backplate.