ACES and Bands

[jan1]

I was just working on a project where I setup the ACES workflow as usual: Clip, then Unicolor mapping from camera to AP1/ACEScct, then ColorGrade, then ACES ODT to map to Rec709.

I noticed that if I make adjustments in the bands the highlights and whites largely seem to affect the same values in the waveform. I am assuming that the ranges that divide the bands 'highlight' and 'white' are calibrated more to a log color space rather than the AP1 color space and thus not sitting right. Same with the softclip roll-off. 

Is there a fix for that or a way to have the bands react properly? If I replace the ACES Unicolor and ODT with just a camera LUT everything works as expected.

I know there's been lots of discussion to make certain operators colorspace aware. Baselight's 'Base Grade' operator is a great example of this working. But even if it has to be done manual, it would be ok for now.

Edited February 25, 2020 by jan1

[Yoav Raz]

Hi Jan

did you try changing the ranges at the tub after bends?

not so familiar with ACES but that tub changes the way mistika behave at the bends. 
hope it helps Yoav

[jan1]

You mean the ranges in the Color Node. Yes, I did play with that, and it seems to influence. But not sure if it does enough and how to set it to deal with a very different color space, rather than just pivoting with in the range.

[Cristobal Bolaños]

On 2/25/2020 at 9:05 PM, jan1 said:

Is there a fix for that or a way to have the bands react properly? If I replace the ACES Unicolor and ODT with just a camera LUT everything works as expected.

Indeed. ACES has the white point very low. To make it work right, try lowering the white pivot to around 70%. That way you will compress a bit the ranges and get a result more reliable from the perspective of a log rec709 color grading. Actually, this is an issue that comes from the ACES curve.

[jan1]

Played with that. The White Pivot is for Primary though, right? I was more looking at the Bands. Specifically I had some highlights need 100IRE which in Rec709 I could easily use the White band to adjust without impacting the rest too much. But in ACES, nothing will unstick the whites around 95 IRE and pull them down.

When I have a regular Rec709 color node, it looks like by default blacks impact the bottom 10% and the whites the top 10% (as visualized in the width of the slope handle in the range graph. In ACES the blacks appears to impact the bottom 70% and the whites are beyond the top.

[Cristobal Bolaños]

Hi Jan!

I'm sorry for the late reply. I seem to have problems with the mail alarm for the forum posts. 

My bad. Indeed, the white pivot is on the Primaries. If you want a general pivot to act on all the tools, that's what the ranges are for. 

So, try lowering the ranges accordingly to around a 70% for ACES.  

[jan1]

On 2/29/2020 at 8:29 AM, Cristobal Bolaños said:

So, try lowering the ranges accordingly to around a 70% for ACES.  

I tried that and it doesn't work. I was unable to achieve a ranges setting that would allow the highlights to be lowered from their peaks.

The problem is more complex than just a range adjustment. There's a mismatch with colors being in a linear space with controls that are designed for log space. In the case of the Baselight's BaseGrade operator the operator actually internally does color space transforms on both ends to unify the way the data and the controls interact regardless of color space. A comparison would be to surround the color node with a 1D LUT on both sides to temporarily change the gamma of the data for the controls.

While ACES has a lot of great advantages, it does make it hard for some of the current tools to work with it. I do see a lot of people work in scene referred work by simply unifying on the Arri LogC color space instead. Maybe that's a better way to go for now.

[Cristobal Bolaños]

Hi Jan!

Sorry for the late reply. I've been digging into this matter. Indeed, it does not have a simple solution, although some situation when working with ACES the ranges can be of well use. 

I've talked to some colorists, and the overall conclusion is that when working with ACES there's no much of a use of the Bands tab, specially when tryng to make a difference between the highlights and whites, due to the ACES curve, in which the white point is too low

 

On 2/29/2020 at 2:55 PM, jan1 said:

While ACES has a lot of great advantages, it does make it hard for some of the current tools to work with it. I do see a lot of people work in scene referred work by simply unifying on the Arri LogC color space instead. Maybe that's a better way to go for now.

Indeed, a lot of people is more familiarized with the Arri LogC color space. That's when the Unicolor gets really handy, when you get used to the modificaciones of one particualr signal, and you can easily turn all your media into that color space or curve.  

[jeff@dungeonbeach.com]

Maybe what's needed is a ColorGradeACES FX which is totally separate from the regular Color Grade node and has the bands scaled correctly + built-in shader/CTL code to fix all the bonkers blue and magenta highlight pixelation that can occur...

[jan1]

Possibly. But that would just create a very fragmented set of tools that's difficult and expensive to maintain. I think a lot can be learned from Baselight who has done a nice job of making the color space journey an integral part of the tool and workflow, rather than some entry/exit point conversion like most other tools. Resolve now offers the CST OFX plugin which people start using heavily, but it is a band-aid, not a properly engineered workflow.

The real answer is to change the platform to be color space aware and then adapt all the tools so they can work in a variety of color spaces with that knowledge. But that's no small feat.

I think we'll see some, like Baselight, lead in this space, and others may technically support ACES, but may not be as practical in that realm for the time being.

Edited March 14, 2020 by jan1