Yeah that TLDR is really messy. It should be ignored. I’m air plane typing and working on some code testing the WGM math right now. And working on another long note from my conversation with Mark the other day.
But there is a point I’m trying to make there. And maybe a better TLDR is I agree with the previous two posts that CAT is mainly useful in ODT. But that I think there are some not-uncommon cases where it would be desirable elsewhere.
And my other main overarching point in this thread is to promote the evaluation of WGM.
Wasn’t trying to pick on your tldr specifically. There’s just lots of misunderstandings of chromatic adaptation out in the wild.
On another note, to @daniele point, is wgm robust to numerically nonsensical XYZ values outside the spectrum locus?
Since all cameras color spaces I’m aware of have D65 white point, my proposal is to add a new color space, AP2, which would have D65 white point and maybe fit into the spectral locus. Or just white point. Or even if it would actually be just a one and only CAT method for use with ACES. Effectively it still would be sort of a new primaries. So no more guessing and errors in pipelines.
ACES 2 will require adjusting pipelines anyway (I mean, AMF). So adopting new working space shouldn’t be a big problem. And AP2 name has digit 2 in it, just like ACES 2. Really the best time for major changes, that will be accepted without any confusion thanks to the naming.
Maybe this all is a bad idea. But both from the forum and private conversations I have a feeling that people who were trying to contribute to developing just stopped because of the lack of the communication between those who a responsible for the big decisions and the community / end users. So I’m just trying to be heard and probably be told, that D60-like white point for ACES has very important reasons to stay, no matter how rarely do people actually deliver D60 masters. Otherwise it looks like just a legacy thing, that this future-oriented framework is trying to keep for no reason.
My point was that if even the developers of Resolve don’t care or don’t understand these things enough to implement it correctly, what to expect from artists? Let’s imagine, I’m a user, who read somewhere that ACES is to make my color management simpler. Just select in and out and you’re done. Yet it’s more like this now: going through ACES code on github to find any info, maybe as comments, on how to choose chromatic adaptation transform for one case or another. Not for a very rare device, but for maybe most common working space today (after camera vendor “native” spaces of course).
Here are the results of a poll I made in August this year. I have a small number of subscribers, so it’s just 83 votes. But still:
(I apologize if the images are too big or too small, I’m on my phone, so not sure about their scale relative to the page)
And as a person, who used to do color grading private master classes, I can say, that even experienced colorists often tell me that they gave up using ACES because they don’t understand it and afraid of it. Gave up before my master classes, not after
Thanks ! This proposal makes sense to me. I have been thinking of a similar colorspace these past years. The ~D60 White Point is one of the reason I have stopped using ACES.
I wonder if this conversation could fit into one of the VWG.
I think the idea of using D65 (as many have commented in this thread) is great. I usually prefer that or D55 for experiments anyway. D55 not for any particular reasons other than it is spectrally close to EE.
But I think as a container space and as a working space including non-pysical primaries has a lot of value in that you can encode all physical colors without using negative values. And in intermediate steps, maybe there are colors outside of Rec2020 which will later be compressed into Rec2020 by the GC/ODT stage.
And for particularly special workflows there are display systems which can display those colors and I think it should be possible to at least contain and encode them (with positive values)
Ignoring convenience, the CIE recommends it for application requiring a daylight illuminant and it is technically the only Standard illuminant of the D series along with A:
All my arguments are based on convenience only. But I will try to ask again, are there any reasons to use D60 for most users or it is just inconvenience with no benefits?
Glad you asked:
D60 produces much nicer skin tones, especially in overexposure, because the bleaching trajectory does not introduce bluish-looking highlights on the skin.
Warmer highlights help skin tones in general, to “stand out” better.
Some argue that D60 is closer to “neutral looking” if you are in a dark room without any other adaptation. I think this is a minor effect when comparing against D65 but huge when comparing to DCI (green) white point. Even after minutes of “trying to establish” a green white point, the observer still does not fully perceive the green white as “colourless”. Some reported a similar (but reduced) effect with D65 in cinema.
On the scene-referred side of things, there is no practical difference as long as you define a single CAT and not many.
I totally agree, that visually D60 looks nicer. But it doesn’t have the same effect on D65 delivery, just by the fact, that there is a CAT for keeping D65 white point, when going to and from ACES color space, that have D60-like white point. At best it ends up to the same result as direct conversion from camera to D65 display. At worst, different CATs in the chain result in something. And even if this 2 CATs difference gives nicer images somehow, it’s still only happens for some cameras, not for all. So there is still a need for standard CAT at least, if changing ACES white point to the same as almost all input sources have, is not an option for an unknown reason.
And all the benefits of D60 white point for any use cases will stay, if we put it into DRT.
I don’t see any downsides to change ACES white point to match source white point, which is D65.
Still all the same benefits for both D60 and D65 masters.
Again, I maybe totally wrong here. I just want to hear, why. For me, as a regular user, for now it sounds like “It is used in film, so it’s cool to have something from film”. For sure this is not a real reason, but the result of the lack of the communication with regular users.
Well if you change the white point of AP0, all your archives will need to be redefined.
If you change the white point, everything associated with AP0 and the spectral definition of ACES is up for grasp. AP0 is an SMPTE standard, so that needs changing too.
All IDTs need to be changed. The house of cards will collapse.
All this hassle just because of the “perceived” inconvenience for some colour space conversions which are not even defined in ACES?
The more straightforward solution is to stick to a single CAT transform going forward, I think.
Sorry for the confusion, I incorrectly called it ACES white point. I wasn’t talking about AP0, but more about ACES working space.
From my point of view it should be a one standard CAT for AP0 and D65 white point for a working space. Which is debatable either it has to contain all the spectral locus colors, or no colors outside of the spectral locus, or something else.
So you would like to have a system that has an encoding space which has a different white point than the working space?
What is the actual problem you are trying to solve here?
I think the underlying issue is that there is more than one CAT method allowed in ACES.
This leads to ambiguous colour space transformation if we go in and out via CSC or IDTs.
Solution: “allow for only one CAT”.
Problem solved.
There may be different reasons for a new working color space, not just white point.
Maybe some color models that only work with colors inside the spectral locus. Or, I don’t know, maybe some models were designed for rec2020 and could somehow be used in DRT developing, but can’t because of some colors that are outside of rec2020.
I don’t know if any of this is true. But I think the upcoming release of ACES 2 is the only time when it’s possible at least to raise this question. And in case of a new working space, D65 white point seems to be a good choice.
From my perspective, AP1 is too small for almost all cameras color spaces. But too big to fit into the spectral locus. I can be wrong here, but isn’t the reason for its slight difference from rec2020 is its D60-like white point?
So no noticeable benefits from being bigger than spectral locus (for example, davinci wide gamut is wider than alexa v3, which is very very convenient for working with maybe the most common camera), but at the same time AP1 is too big to easily deliver rec2020 without the need for gamut mapping in the end of the chain.
One CAT for sure will solve the problem from my first post.
