Using OCIO for color management

White balance means adjusting a certain colour of light to appear achromatic. It’s challenging to even see what the “white” is here, so it’s not entirely the issue, or at least not the more glaring one.

Explain your second point about “dynamic range”. What do you mean?

Well, at some point those “overburn” areas need to get desaturated and faded in a “pleasing way” :smiley:

Ok that is not helping much, I see what’s the problem but, I don’t see such issue whether I shoot canon or sony (super weird color shifting?) It all kinda “burns” naturally towards overexposed areas (at least from what I’ve seen there aren’t any oversaturated colors when overexposed)

Maybe it all lies in some smart color-aware postprocessing tools…

This is a potential solution. The issue is to see the problem clearly.

All commercial DSLRs break in the same way to the best of my knowledge. This is strictly one axis of gamut mapping. There are loosely two to consider. This one is easy to see for most folks on common hardware.

Does it? Have you tested?

The bottom line is that they don’t do what most sane folks would think should do.

That is, we are dealing with two devices, a display and a camera image. For the most part, the display has the intensity range to display the image as it is, but the camera didn’t have enough range to capture the appropriate value.

That is, if we think of a colour in some ABC three channels system, the scene through the ABC filters had a colour coordinate that was too high for the camera to record; the proper colour coordinate lives above and outside the mesh of the camera if you will. The camera could only record one or two of the coordinates, not all three required.

So the question is, given we are device bound, what should happen? Should a DSLR simply offer up the three channels as it saw them, as we have here, skewing the skin tone wildly to a purely saturated yellow, or should something else happen? Eventually, with enough intensity, the camera would display maximum intensity across all three filters, creating an achromatic value, but that doesn’t help the fact that we have yellow skin.

So what should happen?

Well, yes I always judge by display device limitation for “internet purposes” - srgb (or print, up to adobergb) but everything is dictated after raw processing (capture one or rawtherapee) So yeah, I never really deal with raw data as raw as it gets…

As to what should happen, as long as there isn’t a universal way of handling things then the most popular and accepted solution should be adopted… probably :smiley:

There isn’t one. Just the same garbage that renderers, ACES, etc. output.

Ah okay, so it’s back to “bro science” when designer asks for fabric color :smiley:
I’m used to this but it’s kinda weird that something as essential as colors or color space isn’t really standardized to a point where I could just input values without eyeballing stuff and guessing all the possible valuable outcome :smiley: really, a “bro science” experiment this all color stuff in CG

No. The colour transforms are more or less solid. Gamut mapping is missing in this domain. It’s existed in print etc for years, just not yet in this domain.

2 Likes

@lacilaci86 the reason for this color shift is, that in aces. pure colors are not just 1,0,0.

for example, pure red is a 0.61 0.07 0.2 approximately.

If you pick colors in color wheel you also can’t pure result, it is always a mix of something

Not quite. There are pure primaries in ACEScg as well. The same problems are exactly present as discussed above. There is no real nuanced gamut mapping.

At least the color shift seems to be because of this. Of course if you use rgb node to just output pure red you won’t get this shift as well

Did a test in Clarisse and here it seems to be working as expected

i just found this interesting paper about HDR encodings with wide color gamut

edit,and since Blender gets support from AMD,maybe they can help for solutions.

1 Like

What is the expected result? I only see the non-overexposed colors in your picture. The point/problem is what happens with those colors when they become overexposed. For example you have a nice and perfectly working material in one lighting condition but it gets weird colors and saturation once it gets in bright sun etc…

There needs to be some sort of “nice looking” and well controled desaturation and contrast controls for limited bitdepth output. Of course you can use postproduction on linear output outside blender but that’s not a “nice” solution to the problem…

1 Like

How do to it is not the issue, as it is a well researched subject. It’s that ACES and the rest do not.

hm and how was it working in other programms since ACES is widely used?

here a same topic with Blender
https://www.toodee.de/?page_id=1221

and this files for wide gamut and ocio (dont know if they still needed)

ACES isn’t quite used as widely as folks think. But needless to say, ACES has precisely the issues outlined; lack of a more sophisticated gamut mapping. Both high intensity and wider gamut chromaticities “down gamut” will reveal quite a few nasty issues.

ok understand.

at the same site above, the author described how he has setup blender 2.8 to aces 1.1 .maybe usefull for someone.
https://www.toodee.de/?page_id=1720

How do I change the color space of Blender?
I want to leave the one that comes by default and if I need aces, select it
Thx
ColorManag

The videos posted in this thread will tell you how to install ACES for Blender.

You won’t be able to switch between ACES and Filmic while Blender is running, tho. You’ll have to swap out files on your drive and re-launch Blender each time.

I posted a tweak in this thread to make switching between systems slightly easier.

2 Likes