Some concerns

colorcorrection
workflow
Tags: #<Tag:0x00007f9649a74928> #<Tag:0x00007f9649a747c0>

(Charles Boileau) #1

I can post again! @nick finally here is my post…

First off I’d like to say that Grading in ACEScc has took me to another level!! Great stuff!!!

I have two concerns though.

My first concern is with sharing looks with VFXs vendors. The LUT workflow worked pretty well. But now, I’m not totally sure how this should work. I have tested the same workflow in ACES but the specified ACEScc or ACEScg is more tedious and I’m not totally sure a LUT in ACEScg is 100% accurate. Plus, creating a CTL from a LUT is “impossible” from what I hear. Ultimately, the best workflow would be to have a way to convert a LUT (in Resolve) from ACEScc to ACEScg. I think this would require a 2 part CSP LUT that could have a shaperlut to go from ACEScg to ACEScc.

Would anyone have some insight on this workflow? I’m currently working on a feature and would be great to have a good way to do this.

My other concern is versions… will we ever have to specify that to other vendors? Do I now have I specify that I’m working in 1.0.2. Is this something that we’ll have to specify systematically at some point??

Thanks


(Nick Shaw) #2

The simplest approach is to replicate your LUT workflow, specifying the colour space that the LUT is to be applied in to your VFX vendor. If you are grading in ACEScc, you supply a LUT which is your grade only (no IDT, ODT or conversion to/from ACEScc) and if your VFX vendor applies that LUT in ACEScc, with the same ODT you used, then they will see the same thing you did. And if their monitors are different to yours (perhaps sRGB instead of Rec.709) they can just use the appropriate ODT.

If you grade in ACEScct (I would personally recommend you try that) you need to ensure the VFX vendor is using a system which can support ACEScct (the ACES 1.0.3 OCIO config is now available, so systems, such as Nuke, which use OCIO can work in ACEScct). The OCIO File Transform node in Nuke allows you to select the colour space the LUT is applied in.

You should not use a straight cube LUT in ACEScg (or any other linear space). A shaper LUT or mathematical pre-transform is needed to map the input to a suitable mesh spacing. The Academy defines a “Common LUT Format” which is a more flexible LUT format, and can include input and output 1D transforms, 3x3 matrices, ASC CDL as well as cubes. Currently though, that is not generally supported by grading or compositing systems, so is a long term aim, rather than a current practical solution, IMHO.

You may want to read this article on Mixing Light, which explains the VFX workflow for ACES.

One issue you may encounter, and which is mentioned in the article, is that Resolve bakes the RRT and ODT into exported LUTs. Even if you select “No ODT” a transform back to ACES 2065-1 gets baked in. You can counter this by adding an ACES 2065-1 to ACEScc transform to the end of the grade (to invert the ACEScc to ACES 2065-1 transform which will be applied) before exporting your LUT. Pardon the self-promotion, but it may be of interest that I sell a set of ACES DCTL transform files, which can do this.

In terms of specifying to vendors, it is probably safest to specify your ACES version, while things are still changing, and older ACES version my be in circulation. In the long term it should hopefully not be. The RRT is the same in 1.0, 1.0.2 and 1.0.3, so the “look” of ACES doesn’t change.


(Charles Boileau) #3

Thanks for the feedback Nick!

I’ve been experimenting with the workflow and found that it might be possible to switch to YRGB (in Resolve) and then export just the grade as a LUT… Would that work?

Also, why do you recommend working in ACEScct?

I’m also wondering what would be the best way to provide on set LUTs for DPs using ACEScc… It seems like exporting the LUT in ACEScc would most likely be the best way to do this. (as this includes the IDT, RRT, ODT trio).

Thanks!


(Nick Shaw) #4

[quote=“CharlesBoileau, post:3, topic:498”]
I’ve been experimenting with the workflow and found that it might be possible to switch to YRGB (in Resolve) and then export just the grade as a LUT… Would that work?[/quote]
Funnily enough, the same thing occurred to me earlier today. I am not 100% sure that Resolve behaves identically in different modes though. You may find some operations have different maths in YRGB and ACES.

I prefer the way the shadows respond to grading, and find I can adjust them in a more nuanced way. But I am a techie, not a colourist. Use what you find works for you.

Rolling everything into a single LUT is indeed probably the best approach if you just want to load a LUT into a LUT box, rather than have a DIT with an ACES capable system on set.


(Nick Shaw) #5

Mixing Light have an addition to their series of articles on ACES coming out soon. I have written some material for it on the difference between ACEScct and ACEScc. One analogy I made is this:

The linear part of the ACEScct curve near black means it responds to grading in a way which “feels” more familiar to colourists, being more like other log curves. The most obvious way this shows is how with ACEScc, the blacks may feel as if they are “stuck” to the zero line on the waveform. If you try to raise the blacks using the Lift control you will notice that the waveform looks rather like you are peeling bubblegum off the floor – stretching away from it, but with the bottom remaining stuck to it for quite some time, and gradually tearing away.

Try it for yourself. Flip between ACEScct and ACEScc and look at the shadows on the waveform as you move the lift control.


(Peter) #6

@nick: You say the OCIO configuration for 1.0.3 is available?

Where can I find it? The download on http://opencolorio.org/ is still “aces_1.0.1”

Thanks!

Peter


(Steve Tobenkin) #7

Hi Peter,

you can find them here:

also, this thread might interest you:


(Nick Shaw) #8

Steve, it might be worth putting a link to the OCIO Config in the resources section. The thread being under Site Feedback does not make it immediately obvious! You posted your reply while I was still trying to find the thread on here, to give Peter the link.


(Thomas Mansencal) #9

Please be aware that the configuration is 99.9% there but not tagged as final yet, it will only be when merged in Sony repository, there were still some adjustments being made last week like such as role assignments and I think @hpduiker is giving a bit of time for people feedback before opening the PR on origin.

Cheers,

Thomas


(Peter) #10

Thanks!

So should we wait until it’s official?

Peter


(Nick Shaw) #11

Up to you. You can use it, but with sensible caution. My testing has shown no issues.

As Thomas said, all that is likely to change is roles. If you are aware of that, and override manually if necessary, there are unlikely to be problems.


(Thomas Mansencal) #12

As @nick underlined it, provided you are aware of the fact it is not final you should be fine, I think we are at a state where details are getting polished, but devil is in the details!