VRay Sun, Sky and Physical Camera video tut + Bonus Script!

[helpdesk]

O.k. Great...That gives that an answer!
On another note; If you are starting an exterior scene, you use that .255 float value on colors and materials as a starting point and tweak from there right? And you also start by dividing the ISO 200 by 385 for a beginning value in the ISO field? Is this only when sun is positioned directly perpendicular from the ground plane?

[studioDIM]

If you start with a 0.255 multiplier on the materials, that will already bring down the lighting of the sun to ~385x0.255 (for a supposedly white diffuse and a perpendicular sun).
Using a Physcam at its defaults will further bring it down (from 385 to 29, that times the 0.255 above).
So to know by how much to divide the ISOs for, in the case of a white material, you'd have 200/(29x0.255), or 200/7 roughly.
That ~7 should be what you can read in the float values when colorpicking through the VRay vfb.
Unfortunately, the max VFB has issues with float values, as it clamps them at 1.000, whle the non-clamped channel (when saving the render to a float format) seems to disregard the VRay camera exposure.

All in all, i'd suggest you to stick with the VRay VFB if you plan to do the exposure tricks with an accurate measurement of the rendered pixels values...

Lele

[Micha]

Here my test with an interior - my result: don't use the 0.25 method, because this method cause:

(1) wrong relation between light intensity - diffuse reflection - reflection
(2) wrong bounced indirect light intensity
(3) problems with high contrast between direct light and indirect light cause more noise

(1) Problem reflections: look at the blury reflections of the chrom legs at the ground - to strong, because the reflection intensity of the sky is four times higher. If I would set a lower reflection intensity, than the sky would be less visible - ok - but the whole room would be reflected to less too.

(2) look at the shadowed areas like the wall under the window - to dark, not enough indirect light is visible. If we multply the diffuse color with 0.25, than we get max. 25% indirect light reflected - this is not enough.

Lele, what is the problem you try to answer with this method? As an ex maxwell user I think: try to keep it simple and so physical correct as possible. If we find problems, than we should try to find tempory work around and clean bug reports for Vlado. What is the problem? Is the physical sky to dark? The physical sky works with gamma 1. So, the sky at a gamma 2.2 computer screen looks to dark. Maybe this is a reason for the problem. Vray for Rhino dosn't support the physical sky yet, so i can not test it.
Is the indirect reflected light to bright? A multiplier 0.8 should correct. If somebody like to get less, than the simple secondary engine multiplier do a good job.

[studioDIM]

glad you sorted this for us, mate

Lele

[DaForce]

Mmmm I can taste the sarcasm on that one Lele

Micha:
It was Vlado himself that said having both the primary and secondary bounces set to 1.0 is the physically correct way to do things, why do you think they should be changed?

Both the primary and secondary mutlipliers are 1.0 in my images, since this is the physically correct value.

Also what are you basing the white reflects no more than 80% comment from?

[ene.xis]

micha...
maybe ur strong maxwellian past is "blindfooling" u...
i totaly disagree with ur assumptions...

nevertheless hek if it does the trick for u, fine by me...but rn t u taking a little bit too far away by imposing ur 80% and .8 phisical assumptions?

further more...ur using one scene as an example...

i speak for myself but lele s workflow seems quite accurate it s somehow close to mine not-very-scientificall-nor-mathemathical-aproach ()!

[Micha]

@DaForce: I recommend to use the GI multiplier, if somebody like to get less indirect light in a scene. It has nothing todo with Vlados recommendation for physical correct options.
You ask for the 80% issu. It's a dicussed (Maxwell forum) value for typical standard materials like wall colors. It's possibel to get more diffuse reflectance, but it's not seen often. What to do you think? Do you use 100% white for interior wall colors?

Here an image from a Maxwell thread:

If we would multiply all colors with 0.255, than we get a max reflectance of 25%.

@Lele: Sorry, if my words are to harsh. Please tell more, what are the problems you try to solve with your method. Also what you think about the problems (1)..(3)?
I would prefer a workflow that works in every case. What are the effects of the 0.255 method? I would say:
- less indirect light
- sky and emitters show a four times stronger direct visible or reflected color. That cause a wrong ratio between specular reflection and diffuse reflection (diffuse colors).

@ene.xis: I don't understand you. Leles method strong affect the indirect light. What is the problem to take an interior scene? Here it cause problems. What is the goal of the 0.255 method?

[studioDIM]

http://www.chaosgroup.com/forum/phpB...hlight=#152223

The tutorial has no pretense to be "physically correct"
We have no brdf/bsdf implemented in vray (yet? ), so mine is an empirical method to get the exterior shot in range ( i implicitly consider sun and sky at a value of 1 o be "correct" and expose the camera and materials to it).
I am not sure about the reflectance values you state (namely white being 80% reflective).
Reading here , for instance, seems to be closer to what the empirical method and the proposed workflow suggested:



These spheres have an albedo (in other words, "The proportion of incoming light reflected") of 0.1-0.3-0.1 in RGB for a "dark" green colour.
Under these circumstances, a much whiter material may be around 0.3-0.5 in all of them.
Of course, it's scaled down to 8bit, so there may be three ways to produce a white rendered pixel: brighten the material, brighten the lighting (like in the case of the picture above from the top to the bottom row), raise exposure.
Perception matters here, as we end up displaying on 8 bit devices.

A sidekick of darkening materials is that you gain contrast, brighten them, and you lose it.

Further to what i said, i made clear in the tutorial that we want near to no waste of HDR data (ie. pixel colours much above 1.0), as a target for the exercise.
You will notice that the bottom sphere in the left column is PERCEPTUALLY much brighter than the one above, while having the same albedo (hint: different lighting, or different exposure?).
0-255 is limited in both scope and range.
Moving on to float values adds flexibility (Goal #1).
Choosing an arbitrary (ie. i "liked" it, hence it's chosen.) value is only there to aid the psychological transition from the purely 8bit realm to the float one (Goal #2)
Being able to expose through the physcam (assumed "physically correct", as well) in my own eyes need a radically different approach (Goal #3), as a number of situations can lead to very similar, yet intrinsically different, final pixel values.
I propose ONE possible method (UberGoal #1), shown to be working for that particular shot, under those lighting conditions, and ask the user to find the right value for their scenes.
About the "measured" reflectance values (pd, in the linked paper above) it's all down to the way one measures them.
Under what lighting (wavelengths) conditions one measures, how the normalisation is done to graph values, and so on.
Something that paper sketch from the maxwell forum doesn't explain, in the graph, or in the thread for that matter.
Further, try and use your imagination: look at the sun, and see how bright it is (the eye cannot expose it).
If a white wall reflected 80% of that light, do you think the loss of a mere 20% would allow you to see (expose) it as easily as you normally do?
I think not, personally, but was never in a condition (direct or indirect) to measure this with a degree of accuracy.
Maxwell claims it's a spectral renderer, VRay doesn't.
Maxwell uses a closed system, and it's own translator plugins towards Studio, VRay works with most of the max materials and shaders, on top of having just introduced new tools which are proprietary, from within the host application (be it rhino or max or sketchup).
We use version 1.5RC3 under 3dStudio MAX, not 1.49.73 under Rhino.
We have sun and sky and physical camera able to EXPOSE.
The rhino version lacks these three options, do you realise it?

My dry comment is due to the last fact: you seem to have glossed entirely over the fact that we expose a picture with a physical camera, lit by means the Rhino version you use hasn't yet got.
So you unbuckle, and quite strongly, a theory with the exact complementary workflow we all used to have when we lacked the physical camera and sun and sky.
You are entirely correct in whites being 80% (220 white) when working with traditional lighting methods (or with a different rendering engine, for what i know).
That's what i used before these new tools came about (hey i too have some slight experience in rendering images...).
That's also precisely why i started thinking of writing a tutorial.
Hence the title, and the usage of tools i gather you never experimented with, throughout the three parts of the tutorial.
I would suggest you to grab a copy of the same tools we have, and then come back with pertinent critiques to it.
As so far you're criticising something different from what my tutorial is about.
And that's some annoying slippage, mate.

Lele

[percydaman]

Ya I was a little curious why Micha was comparing vastly different Vray builds, when your approach can't be really done WITHOUT the functions included in the latest builds.

[ene.xis]

@ene.xis: I don't understand you. Leles method strong affect the indirect light. What is the problem to take an interior scene? Here it cause problems. What is the goal of the 0.255 method?

well probably lwf makes no sense to u either?
cause if u have that in mind u ll run across color correction...and u might (guess what) ...actually understand the point in darkening ur surfaces...

as in real life...but thats another world out there...ain t it?
or uf doubt it just go out in a cloudy day look for a shaded place take some pictures and judge for urself!

[ene.xis]

phew just finish reading ur post lele!
i cant stop learning with u can i?
well said!

...oh and i agree!

[studioDIM]

A true ex-WoWaholic.
Everything deleted, cd's destroyed.
You can do it Lele!!!

Lol thanks Percy!
I started by interrupting the billing. as of the 5th of january i won't be able to play it anymore. After six more months of rehab, i will have my chars destroyed...
Will definitely have to clean myself up before the expansion comes out :P
Besides, i achieved what i wanted: my feral druid with no epic gear tops the DPS charts in 10men raids
And that is something fairly uncommon (british understatement).
Can be happy with having understood the game mechanics in my very own way.
A good moment to drop the thing, indeed.
End of the OT

Ene.xis, stop that! I'm messing myself with all those compliments.

Lele

[Micha]

I try to understand the method and the effects to the scene. I don't understand why a physical sky and a physical camera change the princip of indirect light transport. But the next SR for Rhino will bring it ... .

If I understand you right, you try a kind of color mapping: 16bit <-> 8bit or? (Goal 1 and 2)

Further, try and use your imagination: look at the sun, and see how bright it is (the eye cannot expose it).
If a white wall reflected 80% of that light, do you think the loss of a mere 20% would allow you to see (expose) it as easily as you normally do?

This is wrong. The sun send the light to the wall from one single sun direction. The wall reflect the light diffuse to all direction of a hemisphere around each surface point. My eye catch only a little part of the diffuse reflected light. So, it's no problem for my eye to see this light. If you say, the wall reflect 100% of a sun ray to my eye, than the wall is a mirror - this is a very seldom case.

If I understand you right, Vray RC3 use a different method of light transport than 1.49.73. At RC3 you can set a color at 25% and the indirect light bounce is not affected? Simple question: if a light ray bounce three times, how much intensity will visible at the end? For 80% reflectance I get final ~50% and I think with 25% material reflectance I get ~2%. You mean this can be ignored or RC3 handle it other?

[percydaman]

2 epicced out characters, 1 a druid, and 1 a warlock.

*sniff, Ill miss laughing at rogues when they try and stunlock a bear with 6.5k of health and 12k of armor, and laugh as they get deathcoiled and kited by my lock, while they're health drains away.... hehehe

[studioDIM]

Further, try and use your imagination: look at the sun, and see how bright it is (the eye cannot expose it).
If a white wall reflected 80% of that light, do you think the loss of a mere 20% would allow you to see (expose) it as easily as you normally do?

This is wrong. The sun send the light to the wall from one single sun direction. The wall reflect the light diffuse to all direction of a hemisphere around each surface point. My eye catch only a little part of the diffuse reflected light. So, it's no problem for my eye to see this light. If you say, the wall reflect 100% of a sun ray to my eye, than the wall is a mirror - this is a very seldom case.

You're right, i forgot to define a context there.
But in the case of CG, when the light is perfectly perpendicular to a flat, white diffuse plane at 80% reflectance (80% white), there's no real scattering happening (mo measured BRDF implemented in VRay).
If the incident light is at 385.0 intensity, it will be bounced back EXACTLY at 80% of that intensity across the whole surface.
Also there's normally no medium between the source, the object and the observer to further scatter and dim the light.
As i keep saying, in CG what looks good is what works well, not what it should be, or how it should work.
For float i mean 32bit values, not 16 bit.
Unbound numbers, in other words, rather than a fixed amount of possible shades.
I chose 0.255 because it RESEMBLES 255 (a number we all know from the past), and it's hence easier to work with. Plus, despite what you may think, it actually works pretty well in treating the image and bounced light.
Rewatch the tutorials, if you don't understand what changes using sky and sun simulated the way they are.
I wrap the whole video around what changes and what needs to be done to somehow cater for these changes.
I'm no maxwellian, and do not believe one ounce of the whole marketing BS about physically correct (until i find a renderer that takes quanta into account, speed of the observer versus speed of light for red/blueshift, and so on and so forth).
I trust in renders that RESEMBLE the physical world believably enough, i know they are a simplified system, with precise bounds and limitations (read post above about the 8 bit display of imagery...).

Lele