LPE Issue Vray6.1 (and6.08) for 3dsMax2023

[jordan_dubosc]

HI!

I Have several issues with LPE I would like to clarify.

1. Weird Expression to get a proper result in Nuke (Garbage/rest)
2. Standard directional VS New VrayLights
3. Full Brut force vs BrutForceLightCache and related issues.

1 - Weird Expression to get a proper result in Nuke (Garbage/rest)

For now in our workflow we are using LightSelect in presets modes, but we have to do some operations in Nuke to get a proper BackToBeauty.

(Full)-(IndirectDiffuse)-(IndirectSpecular)-(DirectDiffuse)-(DirectSpecular)

By doing this we get what's missing between our render and the BacktoBeauty.

We want to use LPE for a full control in Nuke without these operations to optimze the comp process.

The thing is, if I use :

- Direct Diffuse
- Direct Specular
- Indirect Diffuse
- Indirect Specular
- Subsurface (this new one solve parts of the problem)

No matter if I use presetes or custom I am still missing some informations in the glassMaterial:


Full



To get the missing part I have do with LPE pretty much what we do in Nuke with this expression:

(C.*L)-(C<RD>+L)-(C<R[GS]>+L)-(C<RD>.+L)-(C<R[GS]>.+L)-(C<TD>.+L)



How could you explain that? Are we missing something?


2 - Standard directional lights VS New VrayLights

I came across some weird results in Vray6.08 and Vray6.1 using Standard Directional (which is a big part of our process).

You don’t have the same result between the preset DirectDiffuse and the expression for DirectDiffuse:

With the preset every lights are include in the LightSelect

With the custom Expression Standard Directional aren’t taken into account in the LightSelect.

Our lighting team is currently testing the new VrayLight options to see if we can use it in production, in that case it would kinda solve the problem.



3 - Full Brut Force Vs BrutForce LightCache and related issues.

I’ve seen in one of your post https://forums.chaos.com/forum/v-ray...t-select-query that going Full BrutForce would solve the problem which is true after trying it.

Do you have any news on this subject (Internal bug tracker : VMAYA-11778 )

Just to mention again we are working with Vray6.08 (I am currently testing Vray6.1) for 3dsmax 2023 and have the same result.

We can’t really use anything mentioned in this post. Our pipeline is well oriented towards LightCache for now, and we can’t change it atm.



Jordan

[hermit.crab]

1. You're missing the Refraction, which is not a light select preset (obviously). You should get identical results with:

C<T[GS]>.+L

2. Yes, that seems to be the case. Tested a few versions back and it never worked. I've logged the issue (internal bug-tracker id: VMAX-13048 ) for developer investigation. You better use VRayLights or the LightSelect presets instead.
3. No news, I'm afraid. The differences are still present when using LC. I'll bump the issue's priority, however.

Otherwise, are there a lot of very specific LPEs that you're extracting so that you're willing to use LPEs only? Maybe you could optimize your workflow and use the existing elements/presets?

[jordan_dubosc]

Hey! Thank you for the answer!

Yep we didn’t use the refraction but we definitely get the same result with this one (great!).

No we don’t extract very specific LPEs, I was using only LPEs just to make sure we have a similar result for each one of them (maybe you can confirm that there is no problem using preset + LPEs?), that lead me to the following post:

We tried 2 approaches:

1 - Full LPE setup as my previous post (Working on Vray 6.1 with 3dsMax2023)

I don’t have the same Gi (what we said in the previous post, it’s probably linked to the lightcache).

Standard directional not supported.

Appears to be heavier in render time.

2 - We used a Full Preset approach

we don’t have Gi inconsistency anymore

Standard directional are supported

But, there is a difference between the SSS LPE and the Subsurface preset (I assume they should be the same but correct me if I’m wrong).




The result in Subsurface to me is some kind of indirect specular.

(The shader has refraction and translucency. I'll link the scene).

Last but not least talking about light, we tried the same process for Gi from SelfIllumination which is created back in nuke in our actual workflow by taking the beauty pass and subtracting every Lightselect + SelfIllu renderElement.

When using Gi from SI LPE (C<RD>.*O) we get some information but nothing compared to the result we have in nuke.
Also we lost the SelfIllumination objects inside reflection:




When we do a back to beauty with our process we have exactly the beauty rendered.

When we do it with LPEs we are still missing some information.

I’ll link the 2 MaxScene using Vray 6.1 with 3dsMax2023 and both approaches

WorkflowLPE_Presets where I used all the presets available + LPE for the others
WorkflowLPE_Expressions where I used only LPE

And the comp with both back to beauty with Nuke 13.2v4


https://drive.google.com/file/d/1CwC...usp=share_link

(Tell me if you can't acces the files)

Thanks,

Jordan.

[hermit.crab]

The same thing for both approaches:

1. Avoid using shaders, the reflection/refraction of which affects "all channels". This will create double reflections/refractions and may be hard to separate.
2. Use Brute Force to have identical RGB / Comp as already mentioned.
3. Yes, the reflection and SSS of self-illuminated objects are not included in the preset. You may:
a) Capture everything Self-Illumination related -> C.*O
b) Capture GI from SI, SI in Reflections, SI in SSS, and SI. You are missing the reflections and SSS ones -> C<R[GS]>.*O and C<TD>.*O

After that RGB and Comp should be identical.

[jordan_dubosc]

Hey

1. We use "all channels" only in refraction because we render our images without skies in environment nor Domelight, unless there is a way of getting proper refraction alphas to replace the sky or background properly in comp we have no choice. (it was my bad on the glass shader btw).

2-3. After testing, yes it works properly when using BF/BF. We tried both options a/b and it works fine too (with BF/BF and shader on AllChannels). To be honnest I have no idea what would imply working in BF/BF instead of BF/LC (historically I think BF is more accurate but heavier in render time?)

What do you think of changing a full animation pipeline to BF/BF with really heavy scene? (This is our latest movie currently in theaters EPIC TALES) maybe we can talk about this somewhere else?

[hermit.crab]

1. Okay. Just bear in mind that the info will be duplicated in other channels.
2. From the Maya docs (we'll add it in the 3ds Max docs too): "Light Path Expressions that capture specific events after the second GI bounce will work correctly only with Brute Force as a secondary GI engine". LC is more of an approximation for optimization purposes whereas in Brute Force the traced ray is "literal". So it makes sense that LPEs are not perfect with LC (they are "literal" as well).