RawGlobalIllumination * DiffuseFilter do not match GI elements.

[joconnell]

Was it a shader specific issue or a gamma / colour mapping issue? Was thinking with the likes of the hand made raw light passes that dividing by diffuse filter and multiplying by the same diffuse filter should cancel the two out (save for tiny rounding errors) - however would making large colour corrections cause any kind of edge issues?

Good to hear you're sorted!

[Deflaminis]

Can anyone recommend any books or sites about this? I understand "the formula" but I wish I knew more about whats going on underneath to come up with this end result.

[joconnell]

There's a few principles that you need to understand and then you're fine.

Rule 1: Any pass that's driven by light is added.

So for example the vray lighting, Global illumination, specular, self illumination, SSS and reflection passes. For example if you had two lights with a multiplier of 1 in a scene pointed at the same object, the object would get 1 from each light for a sum of 2.

If you wanted to do a really simple vray multipass comp you could get away with the following:

Global illumination (indirect light)
Lighting (direct light)
Specular
Reflection
Refraction

And then add all of those on top of each other. Want less bounce light? Fine, either colour correct the GI pass darker (so it adds less) or drop down it's opacity. Same with any of the other elements.

This is maybe a bit blunt however. Say for example you want to change the colour of the GI light that was hitting your scene without also affecting all of your material colours, or say for example you're doing some kind of car configurator where you'd like to be able to change the colour of a car without affecting it's lighting, then the raw passes start to come in. What happens with the raw passes is they separate the two parts that make up the final pass - in the case of the lighting pass for example, it's the colour of the objects being lit, and then the amount of light that actually hits them.

In the help docs they've examples of each so if we take the lighting pass - http://www.spot3d.com/vray/help/200R...hting.0000.png you've got two parts making this. One is the colours of the objects, two is the amount of light hitting each object. If we break the two apart, you get the flat or diffuse colours of the objects - http://www.spot3d.com/vray/help/200R...ffuse.0000.png and then the amount of light hitting each object - http://www.spot3d.com/vray/help/200R...light.0000.png - If you were to multiply one by the other, they add up to the regular lighting pass. Big deal says you, but what's nice about this is it allows you to get that totally flat colour pass, and then change it's diffuse colour to any other value but still keep your lighting intact. Very handy if you have to do 20 colour variations of the same product.

What's also very powerful is the raw reflection pass - unfortunately spot 3d doesn't have an example pic of this. In your normal reflection element, you get the final reflections on each object in your scene, but say for example the reflections are very weak or even non existent in certain areas of an object, you'll end up getting black in that section of the element. If you try to add more reflection by brightening the reflection element you'll end up with gaps where there was nothing to work with. With the raw reflection element, you'll get a 100% reflection for every object in your scene that had any kind of reflection amount in it's material, almost as if you'd put a mirror material on it. You then also have a second pass called reflection filter, and the reflection filter is a bit like a photoshop layer mask which will dial down the chrome pass to the final amount in your main render. The nice thing about this is that you've got tonnes of reflection information to work with now, and you can colour correct the reflection filter pass to control how your reflection falls off over the surface of an object - it's a bit like having a fresnel option in your compositor to get things just right.

For compositing when you're putting an object into a scene this comes in really handy since you might not get the colour of all the light sources exactly right in your main render, so you can tweak your raw passes for a better match quickly after!

[Richb]

Hi Jo, I'll post a thorough explanation when I get chance, bit busy at work at the minute

It wasn't down to any render settings, gamma issues or compositing methods, basically there was an area behind the sphere in my scene which didn't have any bounced light and no direct light, though it seemed to have some pixel information which was not contained in any of the elements, except for the diffuse. When extracting the Raw passes (clean versions) using the divide method, neither my VrayGI or VrayLight elements could account for this extra bit of information which only the diffuse contained, this left an 'edge' which was very easy to mistake for the typical Halo when compositing with straight raw passes (from the frame buffer).

The testing scene was very basic, clearly too basic as none of our images in-studio ever have an area lacking in both GI and Direct Light, as soon as I began testing with an actual lit scene, I knew the extracted raw pass approach was spot-on, nuke tells me that the composite and the rgb are an exact match down to the pixel - which is awesome!

Hope that helps somebody

I'll post some images which should clarify my point when I get chance

Cheers

[joconnell]

Ah I getcha. What I find happening in a lot of passes is that if I've got an area of very low sampling (say for example a spec pass where the spec is really really weak on a certain object) I've seen some dodgy pixel values - it's like vray early termination has kicked in and you get left with some bonkers colours or values, and often at the edges of objects where for example fresnel reflections are really strong and so there's very little sampling done on the Lighting or GI pass. There might have been something that vray was gaining up when making the diffuse pass? Was it the diffuse filter pass that had the oddities in it?

[Richb]

It was the diffuse Jo, though I have seen some leftover specular info in the lighting pass too but as you say, unless production quality settings are being used, vray can't be expected to sample out all this junk. The main thing I have learned from this test scene is that test scenes should resemble actual production scenes to an extent, and be lit in a similar manner

[joconnell]

True, though it's nice to try things out in simple cases and know something works in theory before relying on it on really heavy production stuff!

[Richb]

Yeah it certainly is, I'm going to try the same test scene again just to put my mind at ease, I'll make sure some light is hitting the whole scene (either directly or indirectly) and then try the comp again, even though I know it will work as expected

[thexdan]

I know this is a very old thread, but I am having very similar issues with raw passes in Vray Next for Houdini. In the end I have gone to using basic GI and Lighting for diffuse as the raw_lighting and raw_gi seem problematic

[^Lele^]

Use non-raw versions off the renderer, and make the raw versions inside the compositing application via division by the respective filters.