RGB 10 black & 180 white

[Leeds3D]

Bumped into a few threads recently about white diffuse values of around 180-200, and blacks at 10 and got a bit lost thinking about the logic as to what that means for other colours inbetween. For example i might be given a jpg image for a clients signage with values of 255 for white, or pale greys that might also be similarly high rgb values. In the same way i've got sample image of brick where the colours are really light greys/beige. I might be given a specific RGB for Costa or Starbucks and if i'm correcting for RGB 10,10,10 for blacks and RGB 180,180,180 for whites whats the best or most appropriate approach to 'correcting' these types of image or fixed RGB values somewhere in the middle? Could i just add an Output Map between my material and all diffuse colours RGBs / JPGs that has the 'Enable Color Map' ticked and use it to clip the top and bottom values perhaps or am i miss understanding whats actually happening here?

[piotrus3333]

output map.

[Ihno]

For just bright textures the rgb level parameter of the output node does the trick. When the texture has very bright and very dark values use the curves. I'd not clip the values. Multiply them down / up so they keep the relationship to each other. The rest is exposure / light to get the colors to look as bright as they should. The max color depends on the material 180 i.e. 0.7 in float might be a little too low for paper and too bright for a sofa fabric. Metals can even have values of 220-250

[joconnell]

I reckon blacks at 10 are even a bit low! Using a colour sampling device on all the "black" things on my desk (mouse mat, keyboard, wacom tablet, iphone screen) are around 50ish!

[Alex_M]

This is all so confusing. Why can't 3ds Max or render engines handle these color transformations internally instead of us having to dial in colors and use textures that look nothing like the ones in real life? If I want white, I want to pick a white color close to 255, Black? I want to pick black. Same with textures - no Output maps with multiplier of 0.7. Why does it have to be so complicated? This 180 white and multiplying textures by 0.7 has been bugging me forever as colors and textures always look mute and darker than they really are in the viewport and material editor. There must be a way to automate this internally if I can do it manually.

[joconnell]

A big problem is our eyes lie to us and cameras are designed to make pretty pictures rather than give us accurate matches of the actual colour of things. I think some of the game engines have some kind of illegal colour check in their diffuse render pass that'll highlight colours that are out of the 50 - 220 range so artists know to change them. It's not the fault of the render engines, we're just feeding them with bad values :/

A good habit is to grab yourself a known colour value (like a mid grey card from a photo shop or a macbeth colour checker) and to take a photo of a surface that you're recreating with the chart beside it - then you'll have a reference point to know if something is darker or lighter than any of the known values on the chart.

[Alex_M]

That's fine. I don't doubt even one bit that our eyes aren't perfect and that these are the "correct" values. My point is why not automate this for the user while showing what our "faulty" would see? It would help so much and save time too.

[joconnell]

Sometimes people want to do unrealistic things where they want to use pure black and white so unfortunately it's really contextual! The renderer is just doing the maths of what you give it so if it was to compress 0 - 255 into 10 - 180, should all the bits in the middle be pushed around too? that's the type of mess you'd deal with. What could be a guide is what some game engines do - use some kind of toggle option to show you colours that are outside of 10-180 (or whatever range you want) so you'd know what to fix first.

The alternative (and what I've been thinking for a while) is to start walking around and randomly colour sampling things in the real world - brick, metal, wood, paint and take a picture of it so we can put up a photo of a thing and the rgb value that it is. At least you'd have a starting point for your data. None of the colour sampling devices are perfect so there's a little bit of inaccuracy (plus or minus 5% I think unless you buy something reeeealy pricey?) but it's better than nothing. Input data accuracy is a huge thing in realism (garbage in / garbage out rule) so it kind of depends on how important it is for your industry and whether you want to start building a reference library for yourself as you need things. Definitely for archvis I reckon there's a few standard things that people use commonly so if you did a few colour samples, you'd be able to make a better guess if other things are darker or lighter than a surface you know the correct value of.

[Morne]

A kind of LUT specifically on the material sample balls would be nice (almost like we're already getting with the 2.2 vs linear thing), but specifically to make the material ball sample BRIGHTER. That way we will automatically make things "darker", but what you see in the material editor will be more correct for your eye. This LUT is then obvisouly ignored during render, to give you more "correct" results. I wonder how much trouble this kind of thing would be, or if it can in fact be some kind of answer... ( vlado )

[joconnell]

A macbeth chart and this software could help to straighten out colours in reference photos - https://gumroad.com/l/calibrate

[squintnic]

I do 30-200 range generally and then post to suit or in VFB adjustments

[Ihno]

A kind of LUT specifically on the material sample balls would be nice (almost like we're already getting with the 2.2 vs linear thing), but specifically to make the material ball sample BRIGHTER. That way we will automatically make things "darker", but what you see in the material editor will be more correct for your eye. This LUT is then obvisouly ignored during render, to give you more "correct" results. I wonder how much trouble this kind of thing would be, or if it can in fact be some kind of answer... ( vlado )

1+
bla bla character limit

[dgruwier]

This has been discussed here, including the closest thing there is to an official response yet:
https://forums.chaosgroup.com/forum/...ainter-in-vray

The problem is that diffuse values depend way too much on context to correct automatically, and picking the right ones at least a minimum of attention and judgement. You need to look at reference images and albedo values to get an idea of what your target is, then make the individual corrections necessary to get your diffuse textures or colors closer to the correct value for that material. Those corrections are going to be completely different from texture to texture, and in some cases (like with pre-calibrated PBR textures) are not even necessary in the first place. Besides, the 10-180 range thing is only ever meant as a general rule of thumb that you can use to sanity check values, and the exact number range is also different depending on who you ask.

Instead what you want are tools that let you take real world measurements and use them as a guideline for adjusting your own diffuse values. More about that in the thread I linked to.


Morne Unless I misunderstand what you're asking for, you don't want to correct the beauty render (which the sample balls are, just a small version) which would affect everything including otherwise correct reflections and self-illuminated sources. You want to correct the values of your diffuse map or color as they go into your material, and if those values are correct to begin with, your sample ball will look correct.

[piotrus3333]

I’m sticking to this range:
RGB (sRGB)
10(59)-230(242) [charcoal-fresh snow]
46(117) photo graycard
204(230) fresh white interior wall

i remember Albedo app for iOS that you use with a small graycard to get material’s reflectance. not the gratest but cheap.

[Ihno]

Im always wondering if those srgb values from the macbeth chart are actually corresponding to their albedo when converted to linear. https://xritephoto.com/ph_product_ov...=5159&catid=28

they say the white patch has a value around 243 in srgb that makes me rather sceptical it seems a bit high doesn't it?