Do you have a link, or is this an AI summary?
Some of this is poorly worded.
Also, the last paragraph is exactly what everyone has been saying, store in AP0 and work in AP1. Also what @dazedinheaven clarified. So did something get lost in translation?
I think this is not up to ‘some online summary’ or ‘some VFX sup’. It standards organization (like the Academy) and delivery and workflow specs from Netflix et al that are the authoritative source and recommendation.
I think the negative values are in reference to the AP0 gamut triangle going into the negative in order to encompass the CIE 1931 arch.
So not negative values in a floating point sense, but rather that because the CIE arch is plotted into an X/Y grid, and because it’s got a funky shape AND because any computer gamut is defined as a triangle of R, G and B, the AP0 triangle has its left corner in the negative.
I’m sorry, but “peace,” and solely for the sake of correct meaning.
I do. Just trying to promote the full gamut of English prose.
Well, that attitude may be exactly why some eager vfx sup may me thinking he’s smarter than the folks at Netflix, if that is indeed the source of this saga.
Less bravado, more information, more careful consideration and verification wins the day.
Don’t blame me for your ignorance.
From the Academy who make ACES who probably know a shitload more about it than you since they created the standard.
The reasoning for ACEScg which completely backs up what I previously said.
ACEScg ― It uses AP1 wide-gamut primaries, is photometrically linear (gamma-1.0) and floating-point arithmetics. This exists to allow artists to visually process any pixel of an ACES-encoded image during compositing, painting and other CG-related processes, but without losing any information to colours that may not be representable for different reasons (including out-of-gamut’s). For example, codvalues <0.0 and >1.0 are expected to represent colours outside the AP1 gamut (yet within AP0’s).
So because you can’t see the extremities of what you are doing in ACES AP0 you could have noise in your comps and renders that you can’t see when you are working in ACES AP0. So when you bring that back into AP1 or P3 then you see the noise that you couldn’t see in comping. Which is what I have been saying. You can get noise or clipping at the edges of the working gamut. I can’t present it any simpler than that. You can also create negative, out of gamut values going back to AP1 for grading after comping in AP0 which presents as noise.
Comping & grading in AP1 is for people in the know who follow the standards of the colour science they are working in, I.e. the ones in the know. AP0 is for those who are doing something different for no discernably good reason. It’s hardly a comparison between men and boys.
Understanding colour management can be a daunting topic for many and as the journey of discovery continues it is important to keep an open mind. Be open to other opinions and workflows.
With this being a tricky topic for many to comprehend, any breakthroughs can feel like significant light bulb 
switching on moments.
Be sure that this doesn’t create a negative effect of closing you off to any contradictory information.
It would not surprise me if some vendors or VFX supervisors insist on workflows that seem suboptimal to most people.
There are some excellent colour management discussions in this forum. Many of us have self educated ourselves on this topic and will defend our opinions passionately.
I find it pretty hilarious that we talk about “encompassing more” “more gamut” something something “not clipped”
Think we first need to understand what this “gamut” even means and how it mathes out.
In general the primaries just tell us “if you have full green 0/1/0 in RGB thats here on XYZ” XYZ then directly reffering to “the human observer” (which is all kinda BS in a way anyhow but whatever).
So lets imagine the full-green color of 2065-1 -
now we define a “smaller” gamut like rec709, can we make any color value relative to rec709 point at the same XYZ coordinate that the full 0/1/0 AP0 primary sits at?
What is “more green” than full rec709 green ? Hell we just subtract red and blue - negative values baby .
So yes - guess what - we can actually describe any color on XYZ using any set of primaries we want to, no problem ist all “relative” we just scale the numbers relative to these 3 abritary primaries.
So by that logic why not use XYZ directly? (hint; DCPs)
Regardless of these theoretical things, i think know why shows force people to use AP0 even tho its not recommended,
Lack of understanding " its bigger so its better" see the above
Wrong scope of reference when talking about QC, and i think thats the biggest one here.
Lets say you ingest your arriraw directly to AP0 , what happens here ? the native camera gamut values are beign plotted on XYZ and then described using AP0 / cool so now these AP0 values are beign described as the “original media” that we compare the future compositing output to.
These can allready have negative values to describe values outside of ap0 that a camera might still be able to encodex
A raw doesnt have any “original picture” its a data-soup just to sidetrack.
Now we take this AP0 lets say 16b float exr and throw it into nuke withba acesCG workingspace what happens →
32bit working space, 16b source so there is a upconversion where the space between samples needs to be filled with something and new values are made up with 32bit precision, so now you have 32bit float data internally, relative to AP1/acesCG.
Now we write out a 16b acesCG precomp , bow our 32bit interpolated values are beign rounded to 16bit precision again.
you load that 16b acesCG file back in and continue compositing , notice we now have a rounding error to our original plate.
you now export a AP0 exr at the end of your comp but you still have a rounding error against the original 16bf AP0 data.
This would easily lead someone to believe “we have to do everything in AP0” as they dont recognize the difference as a rounding error.
However - if you take the AP0 data and then turn it back to camera native you will also have a rounding error there allready, if you would have just gone camera ->ap1 directly and then leave it there you would end up with the same result.
So its a pixel-fudger issue in my opinion due to lack of understanding.
If you really think that nobody should touch “camera encoding” then you should use your cameras native gamut instead, why not use linear/alexawidegamut4 as your exchange format in the first place, (in fact i worked on some shows that did that, snd i heard its becomming more and more popular)
In my personal opinion 2065-1 should be erased from existence. But thats just me, i dont think we need it, in my facility exrs are always acesCG (but i dont do longform anymore but ive been there done that got the qc reports)
Next thing would be to dive in “why” the size of your working space gamut matters, because thats pretty interesting as well (energy ratios and some fun stuff to be found there) .
Some fun things to think about with the math just 2 of the things , if tou consider things like energy conservation and stuff in 3D renders stuff becomes a bit more complex but here are just zw examples i can think of why the math matters and why we “need” a narrower shaped working space :
When you interpolate between two colors (for example, crossfading, soft edges, glows, blur), color values in a massive gamut like AP0 can pass through regions of “nonphysical” or “imaginary” color — resulting in weird hues, saturation shifts, or banding.
Poor Color Grading Response
If you use a huge gamut, simple grading operations (like increasing saturation, pushing midtones warm, etc.) behave very unintuitively. A slight change can cause huge hue shifts or saturation changes because the space is “too loose” — the axes don’t match human perception tightly enough.
This has been one of the sadder threads on this forum in recent memory.
Learning is not free on either the teacher nor the student. Nobody is obligated to make information available to you so you can learn. There’s no free trial or freemium version of trade knowledge.
Despite that many very experienced folks offered their time in this thread to explain and untangle this for you. As the person learning, you need to start with an open mind, take time to absorb the information, cross check it with other information you (as not all information is always correct despite best intentions) and then update your mental model accordingly.
If you go into the conversation with the assumption that others are wrong or don’t know what they’re talking about, it makes it harder see the facts, or realize that your prior understanding may have been incomplete.
We all probably have stories where we assumed A to be correct for a long time, even delivered client work on the assumption of A, and then we learn we had it all backwards. That happens to everyone, no matter their experience. Sometimes that can be embarrassing, to have a junior moment. But I’m always glad for the opportunity to revise my understanding and deliver more professional results going forward.
I’d rather take the ego ding of being proven wrong, then continuing to make a mistake that contributes to an inferior product and worse yet, propagate a flawed understanding.
And yes, we have clients all the time that ask us for things that are incorrect or outright stupid. There is a lot of misinformation out there. As a professional it’s our job to recognize that. If the client is open minded enough to hear an explanation, you try to change their mind. If not, see if you can work around it with integrity. If that’s not possibly you decline the job.
In the case at hand, you would first need to verify if the client asked you to ‘do all work in AP0’ or ‘work with files in AP0’. It’s a nuance that may have gotten lost in translation. More likely as has been pointed out - it was receive and deliver files in AP0, but no need to do the work in that fashion. And even if they did mean that in a moment of misinformation, you could decide to internally work in AP1 based on your professional judgement, and then deliver the files in AP0, and the client should have what they asked for, and none the wiser.
Lastly - we’re not boys and girls here. We’re all seasoned and mature professionals. Insult is not an escape hatch from having traveled down the unwise path.
Looking forward to more productive and enlightening conversations going forwrad.
Aww shucks, sorry to hear you are feeling under the weather.
There’s a vaccine for this illness.
It’s called AP1.
Just got here. What did I miss?