Disable Max Gamma Correction and use VRay's LWF button?

[ARTECONI-CGI]

Hi Vlado,

Thanks a lot for taking the time to reply to all of my questions.
The main concern I have is about the color picker. As you say you would gamma correct a RGB value only when the color comes from an 8 bit texture.
After years of trial and errors I come to the conclusion that I always want to color correct the rgb value, also if it doesn't come from an 8 bit texture because, as it is now, it's very very hard to control.
I understand what you say: In both the cases I am always picking up a color between black and white so ,potentially, I can achieve the same result.
But believe me, when you plug it with .4545 node, what you see and render makes much more sense.
If you don't, all the shader is forced towards the bright areas and you end up with values like this:

0 if you want pure black (and you probably never want pure black)
1-5 darks
5-50 midtones
50-100 brights
100-255 highlights

meanwhile in the .4545 scenario the values are more easily distributed:

0-50 blacks
50-100 darks
100-150 midtones
150-200 bright
200-255 highlights

This is a rough estimate but please give it a try.
Create a material in both the setups and you will find that the .4545 approach is much more easy to setup and control.
While, without it, you will see that when you create a reflective material assigning an RGB value of 50,50,50 gives you a surface already too much reflective.
A black plastic needs RGB values lower than 1,1,1 in reflection to look natural so you end up having to do tricks for those fractional values.
The same happens with the refraction.
The material become immediately way too refractive at 50,50,50 and I have the feeling that this happens with most of the things that you control with an RGB value in Vray.

This is the reason why I think the color picker should be color corrected automatically by default.
This happens in Mray through the gamma and lut panel in Max but doesn't happen with Vray, so, I was guessing that it was due to the fact that the Mray uses floating point value (0 to 1) meanwhile Vray RGB (0-255)

Kind Regards,

Giacomo.

[vlado]

If you don't, all the shader is forced towards the bright areas and you end up with values like this:...

This is true, yes. We could add floating-point color pickers for V-Ray materials if you think it will help. But it's totally outside of your previous questions and you didn't mention it in the previous post.

This is the reason why I think the color picker should be color corrected automatically by default.

I don't agree here... if you wanted to enter 0.5, 0.5, 0.5 (or 128, 128, 128 ) then this value should be used in your final render. I don't think a renderer should have the liberty to go and change that value behind your back unless this was clearly marked in some way. We could, for example, modify the VRayColor texture to do that if you wanted, but I don't think it should be an automatic process.

This happens in Mray through the gamma and lut panel in Max but doesn't happen with Vray

No, I don't think so. If I enter (0.5, 0.5, 0.5) in mental ray with Arch&Design, and (128, 128, 128 ) in V-Ray, I get exactly the same result. Same for e.g. 64, 64, 64 in V-Ray and 0.25, 0.25, 0.25 in mental ray or 192,192,192 and 0.75, 0.75, 0.75. If you have a simple scene that shows that they are different, it will be helpful.

Best regards,
Vlado

[ARTECONI-CGI]

Hi Vlado,

Thanks for the reply.
I don't have a lot of time to run some more tests now , deadline as usual, but I will as soon as I can because this is a topic really important for me.

You are right,I was wrong: a vray grey 128 correspond to a grey 0,5 in mray and so on 64 to 0,25, 192 to 0,75.
I made another test for the reflection and those didn't match, the vray one was brighter but I wan't to doublecheck the test again before claiming something wrong again.

Anyway the feeling I have in general for refraction, fog color, and most of the parameter that can be controlled by an rgb value, is that they are hard to control because the curve applied on top, makes very easy to pick a bright color instead of a dark one.

It seems like that those spinners needs to be gammacorrected with a .4545 node.
I think this is what happen using the LWF button and I think it gives so much easy control to the shader. More photorealistic render and less washed out.

Another clear example is the vray 2sided mat.
Create a lamp and place a light inside. You'll get a much nicer and controllable shader if you use the lwf button (that use a .4545 correction i suppose) than with the traditional workflow.

Same things happens with the fog color. Much more controllable with the lwf button than not.
I'm not promoting mray but I find the that shaders it is much more controllable. As per the LWF button in vray.

Anyway, I'll take the weekend to run some more tests and will get back to the forum.

Kind Regards,

Giacomo.

[vlado]

I made another test for the reflection and those didn't match, the vray one was brighter but I wan't to doublecheck the test again before claiming something wrong again.

My test with reflections matched too.

Anyway the feeling I have in general for refraction, fog color, and most of the parameter that can be controlled by an rgb value, is that they are hard to control because the curve applied on top, makes very easy to pick a bright color instead of a dark one.

This is true, yes. But it should be something that's solved in the color picker, not the renderer. We could write our own color picker that does just that.

It seems like that those spinners needs to be gammacorrected with a .4545 node.

That might be the wrong way to solve the issue. Again, solving this in the color picker is probably best and I think I'm liking this idea more and more.

Another clear example is the vray 2sided mat. Create a lamp and place a light inside. You'll get a much nicer and controllable shader if you use the lwf button (that use a .4545 correction i suppose) than with the traditional workflow.

But the VRay2SidedMtl material is not affected by the LWF option (only its submaterials are).

Same things happens with the fog color. Much more controllable with the lwf button than not.

I don't think the fog color is affected by that option either.

I'm not promoting mray but I find the that shaders it is much more controllable. As per the LWF button in vray.

Again, I would respectfully disagree here They do have the floating point color picker, but that's all.

Best regards,
Vlado

[trixian]

I'm getting the impression this all boils down to the fact that you prefer to manually pick colours in a gradient that is not gamma corrected, thus your preference for how it looked when not using max's gamma and turning on vrays LWF checkbox.
Therefore, as Vlado points out, this is a colour picker issue, and could be solved with a new custom picker.
What I never really understood, is why the gamma correction is so biased towards highlights. The whole point of it is to move the colours so the look more evenly distributed to us puny humans, but AFAICS, they push it too far towards the lighter side, making most of the lighter colours indistinguishable from each other, just as the linear gradient makes darker tones indistinguishable.

The question is then, will this new and improved colour picker be gamma corrected, linear or custom?
Where is the sweet spot where eyeballing a colour in a picker makes the shader behave predictably?

One thing I noticed after getting VrayRT, is interactively tweaking colours for fog and other such parameters while looking at the result gets a hell of a lot easier than looking at first a gradient, then a material swatch, and finally a rendered image.

[ARTECONI-CGI]

Hi guys,

Thanks for the answers.
Again, no spare time for tests but I will, as soon as things slow down.

Vlado, I'm not a developer and my understanding of the math behind the rendering engine is limited so my impressions are completely based on "field" experience.
I have adopted the LWF at the early stage and with the time I have developed the feeling that something is not corrected along the process.
I'm not sure if it is just the color picker or what else.
In general I find that the LWF button method is more photorealistic than the traditional LWF.
Probably because it corrects the color picker. Well, if I'm not wrong, it corrects the entire shader.
Now, I don't know if correcting the full shader is right or wrong.
I understand as well that there is no right or wrong because you can achieve the same result in both cases.
But it is so much easy and more natural to guess the right amount of reflection etc etc etc with the LWF button than the other way round.

Again, there is another doubt abut the traditional LWF. The fresnel button on the reflection / refraction and all the max procedural maps as noise, splat etc etc.
I wonder if they receive the same treatment as the color picker. I mean, do they get brighten up and makes it more difficult to control?
Would they behave better and would be easier to control under a .45 node?
I really don't know what the answer is but I would like to understand what is your point of view about it and what is really happening.

Again, with the LWF button things seems more right to me.
I honestly would stick with it and stop bothering you but, this as well, is not a 100% correct LWF because it doesn't correct some maps. (vraylightmaterial, lights....)
I really wonder: Are we sure that correcting the entire shader is wrong? If so, can you explain why? I know the whole story behind it: jpegs have sRGB embedded and we linearize it... and on and on ....
....but it still looks better with the LWF button on!

About the vray2sided material: the translucency parameter is controlled by a RGB value so I think this will work better if corrected by a .45 node.
Same thing for the fog color: still controlled by a RGB value. It was easier to control it before the advent of the LWF.

Trixian:

"What I never really understood, is why the gamma correction is so biased towards highlights. The whole point of it is to move the colours so the look more evenly distributed to us puny humans, but AFAICS, they push it too far towards the lighter side, making most of the lighter colours indistinguishable from each other, just as the linear gradient makes darker tones indistinguishable."

I agree with what you say above. And I agree with the improvements vrayRT is bringing. The problem is that I work on scenes with over 10 milion polygons. Difficult to run in RT.

Kind Regards,

Giacomo.

[vlado]

In general I find that the LWF button method is more photorealistic than the traditional LWF.

It is a lot less consistent, I will explain this below since you mention Fresnel reflections.

But it is so much easy and more natural to guess the right amount of reflection etc etc etc with the LWF button than the other way round.

As I said, I agree with this. But I'm not entirely convinced that the "Linear workflow" option is the right solution to the problem.

The fresnel button on the reflection / refraction and all the max procedural maps as noise, splat etc etc.

For the Fresnel maps specifically, they are computed in the physical linear color space of the renderer; as such, the maps themselves should not be gamma-corrected. Only their input colors or textures must be corrected as needed, but not their output. However the "Linear workflow" option does not recognize this and blindly corrects anything connected to the reflection slot, which is totally and utterly wrong with a Fresnel map.

Would they behave better and would be easier to control under a .45 node?

It depends; take for example a Gradient map. Without LWF or with the "Linear workflow" option, it appears as a linear gradient from black to white. However if you use it as a texture in the standard 3ds Max workflow, it will appear as the same washed out gradient that you get in the 3ds Max color picker. I can see where the first option is probably more desirable and in line with what the user expects.

Again, with the LWF button things seems more right to me.

In the Gradient map case above, it will indeed look more like you would expect it to. But with a Fresnel map, it will produce a wrong result.

I really wonder: Are we sure that correcting the entire shader is wrong?

In the case of a Fresnel map, it is wrong. In other cases as with the Gradient map, it can be argued which one is "correct" and I don't really know the answer.

If so, can you explain why?

Because it blindly corrects everything, even things that should not be corrected like the Falloff map.

....but it still looks better with the LWF button on!

I'm not arguing with that

About the vray2sided material: the translucency parameter is controlled by a RGB value so I think this will work better if corrected by a .45 node.

I don't know about this one. If I set it to medium grey, I probably want the front and back surface to contribute equally.

Same thing for the fog color: still controlled by a RGB value. It was easier to control it before the advent of the LWF.

I suppose so; but as I said the fog color is not affected by the "Linear workflow" option anyways. You can use the fog multiplier to change the rate at which the fog color affects the result.

"What I never really understood, is why the gamma correction is so biased towards highlights. The whole point of it is to move the colours so the look more evenly distributed to us puny humans, but AFAICS, they push it too far towards the lighter side, making most of the lighter colours indistinguishable from each other, just as the linear gradient makes darker tones indistinguishable."

The whole point of the gamma is to allow you to see the images as they really look like. They look washed out because human perception is not linear; so a gradient that is linear to the renderer, is not linear to people. The purpose of the color picker is to give you an idea of what result you will get for a given input value. I agree that this tends to push the color picker towards white, which is inconvenient if you want to work with dark colors. In that case, a non-linear color picker would work better.

The problem is that I work on scenes with over 10 milion polygons. Difficult to run in RT.

You can still isolate objects and adjust their materials individually; no need to keep the RT rendering an entire building if you are adjusting just the material on a sofa in one of the rooms.

Best regards,
Vlado

[^Lele^]

This is true, yes. We could add floating-point color pickers for V-Ray materials if you think it will help. But it's totally outside of your previous questions and you didn't mention it in the previous post.

I use a vray color texture if i need float RGB input rather than an integer, for whatever procedural reason.
Hardly immediate as a color picker, agreeably, however is a wee bit more flexible (gamma controls, global multipliers, and so on), and i have never found anything but absolute correspondence between the maths of the two methods (ie. 0.5 float is 128 int).
I believe since max2010 color is internally converted in float, no matter the color picker choice of value display.
Before the VrayColor map, i made myself a calibrated (ie. of known value) image, and put it into an output map to drive it with the rgb multiplier.
Or in the case of Fog, use the fog multiplier (thank goodness for the multiplicative nature of the operations...).

It surely could be made more elegant or streamlined (and dangerous. i can see right now whites on diffuse materials at 128 FLOAT for a mid-gray, with the FP color pickers), but it's not like it's missing a thing, right now, structurally, imho.

[ARTECONI-CGI]

Hi all,

Thank you very much for the in-depth analysis Vlado, I really appreciate your effort and I can honestly say this is the best support I have ever received from a company. I wish Autodesk was the same....

I understand that the best approach to the LWF is the traditional one + vraycolor. Thanks Lele as well for your tip.

I will then suggest to introduce an option to replace the standard RGB color picker with the vraycolor and the possibility to have the color already corrected by default.
It would be very helpful.

I will still have to figure out what is visually more correct about the procedural maps, tough..


Vray RT is great and the suggestion is good but I won't have the same lighting in isolation mode.

Thanks again.

Giacomo.

[raid]

HI all,
i was used to LWf1 using Max 2013. I jsut finished a longer project and installed 2014 yesterday and immediatelly saw the problem and found this topic. Apparently i used Method 1 but never color corrected all materials.
I have jsut 2 simple questions since normally i always use use pictures,jpgs, fotographs on a daily basis. Do i really have to add a colorcorrection texture and change the gamma value everywhere? (some scenes have more then 100 stacked materials....) ?
The new mode "color mapping only no gamma" is now switched on, since i am totally confused aswell, what happend to "do not affect colors" and how does that interfere with colorcorrecting every material with the right gamma?
Sorry for asking again but in the erliaer post it was stated to use "dont affect colors" now this is not available....(i guess its now colormapping only now gamma, if i want to use the option).
I am feared of the idea to use everytime the colorcorrection texture on every material slot except fresnel.

[trixian]

Just a clarification, regarding gamma corrected gradients.

What I meant was that for our use in a 3d application, it seems strange that the gradient does not get the appearance of being evenly distributed when corrected, since we need this gradient to make (informed) decisions based on visual input on what colour/value is used for something. The result now is that we rely on past experience and/or rendered previews to chose our values, as opposed to "getting it right" at the first stage when looking at the colour swatch. This is of course a bit exaggerated, as we seldom actually get it right before seeing the rendered result and tweaking.
Should not the correct gamma for "perception" actually make the corrected gradient appear completely evenly distributed?

[wyszolmirski]

This is great thread. Thanks Vlado for clarification. Too bad I didn't have time to read the forums earlier.

Few months ago I was misguided by Chaosgroup support while I was preparing video tutorial about LWF for chaosgroup youtube channel... I was specifically asked to use "Linear workflow" checkbox... at first it seemed weird request but results look OK so I went with it, thinking that something changed about it over years and it was suddenly good to use... Recorded the tutorial and even we've started using this technique in the studio for some time. And now it turns out we've wasted some time because of this miss communication. Anyway I've taken the tutorial down from my vimeo page (now I know why Chaosgroup never released it on the youtube channel) ...

[ARTECONI-CGI]

Trixian, I agree on the doubt about the color swatch....


Wyszolmirski: sadly the same Chaosgroup Forum Youtube channel is packed of tutorial with a wrong LWF setup:

Most of the tutorial you find on the same Chaosgroup Channel doesn't use the correct LWF, you can easily understand from the color swatches that appears darker.

http://www.youtube.com/watch?v=1LX7Mo8Bzr0

http://www.youtube.com/watch?v=DsLGj88RdR0

http://www.youtube.com/watch?v=7ehsaSO2auI

and Gnomon DVDs:

http://www.thegnomonworkshop.com/store/product/987

http://www.thegnomonworkshop.com/store/product/1022/

I can't believe that so many professional top artists doesn't know how to setup a correct LWF.....


Giacomo.

[Deflaminis]

I can't believe that so many professional top artists doesn't know how to setup a correct LWF.....


Giacomo.

I've read this thread through a few times now and it seems to me that there *is* no correct LWF setup. Every implementation has something wrong with it, and it's getting exasperating.

Then I fire up Maxwell and everything just... works. No problems at all, it's getting frustrating.

[vlado]

I've read this thread through a few times now and it seems to me that there *is* no correct LWF setup.

It is true that there are several possible ways to do a linear workflow. They can all be correct.

Every implementation has something wrong with it, and it's getting exasperating.

Nope, they are all correct.

Then I fire up Maxwell and everything just... works. No problems at all, it's getting frustrating.

How do you know that? I don't know of a render engine that can guess 100% what you want to do. Maxwell tries to do its best, but there are cases where it will guess wrong.

Best regards,
Vlado