Color Management in VFX — Overall

In the visual effects industry, there are still lots of artists who feel scared for color management because it’s quite hard to understand at the beginning and sometimes it’s too difficult for them to figure out why the color is not right, so they just continue to do their work in an inaccurate way and try to get closer to the real look. But what if there’s a better way that makes the whole pipeline more efficient and looks more realistic? We do have a solution since ACES appears. But we won’t dig too deep in it, just the overall color management between regular linear workflow and ACES.

But hold on. Before we go into the topic, in case there are still some people not familiar with the basic concept of colorspace, I just do a quick recap.

  • All the concepts here have been simplified to have a picture of the whole idea for the artists to understand easily. I try not to use terms that are too incomprehensible.

What is Colorspace?

There are three important parameters that define the colorspace: Gamma, Gamut and White Point.

Let’s look at the graphs from Cinematic Color 2, gamma is a non-linear operation to encode and decode the illuminance, by taking advantage of the non-linear manner in which humans perceive light and color; Gamut is defined by the three primaries including red, blue and green, which at each top of the triangle in the graph; And we still need to define which point is the brightest and purest white, that we so called the white point.

Every colorspace has its own definition in gamma, gamut and white point for different purposes and scenario like the example in the graphs: sRGB has three primaries at the top of the red triangle which comes from Rec.709 and reflect the approximate color of consumer CRT phosphors; it has been designed to be seen in the dim lits environment that defines the white point as D65, a standard illuminant has color temperature of approximately 6500K. Its non-linear curve is similar to the gamma response of a CRT and efficiently reflects the color of human-discernible light levels.

WHAT IS IT ALL ABOUT in VFX?

If we put a middle grey as 0.5 under the sRGB colorspace, it will be 0.18 in the linear space. And here is the problem, in the linear space everything is calculated linearly like how lights work in the real world. Let’s double the intensity, so 0.18 becomes 0.36 and convert the value back to sRGB colorspace as 0.63. OK, let’s go back to do the same thing: double the intensity for the middle grey 0.5 in sRGB colorspace and we got 2, which obviously is not equal to the value 0.63 calculated from the linear space. The result shows the difference doing the same operation in the linear and sRGB space. This is why we need to linearize all the files with non-linear gamma curves.

Cool. So now we know working in visual effects we need to keep two things in mind:

  1. Linearize everything
  2. Set a proper gamut and make sure every source is interpreted by the same primaries and white point.

DIFFERENT LINEAR WORKFLOW

Now we start with the whole workflow. Before the ACES jump into the game, the basic concept is around keeping the colorspace consistent in every step.

In the traditional workflow, the most commonly used colorspace in VFX is linear-sRGB. From all the assets for CG artists need to linearize the sRGB images before doing computation. It is the option for matte painters to linearize sRGB images or not depending on studios, but I don’t have many knowledge for ACES in matte painting. I will have an other post in other day. In compositing, compers convert log footage back to linear-sRGB then they are able to combine other elements from the CG department and matte painting department in the same linear-sRGB colorspace. That will make everything mathematically correct.

Alright, let’s go into the ACES workflow. In fact all the steps are pretty much the same. The major difference is it maps the sRGB primaries to ACEScg, which allows different color settings in different cameras from their own manufacturers can be converted into the same giant colorspace and artists have better control to manipulate color in the larger gamut with high dynamic range.

sRGB vs ACES?

The limitation for the traditional linear workflow here is the primaries only confined in sRGB gamut, which means pixels from objects in 3D applications receive and respond lights will also be clipped in the sRGB gamut. As the graphs from VFX HARRY’S DESK below, we can see it has no rolloff in the highlight area in linear-sRGB and that will cause overexposure and over-saturation in our images.

Here is the better explanation from Chris Brejon’s website. Chris Brejon is a ACES mentor and he wrote amazing context about CG cinematography. Probably need some patience and time to digest but please check it out.

Cool. So the color model is plotted in CIE-XYZ and flatten to a lotus. the first image is rendered in linear-sRGB. As we mentioned earlier, all the colors have been confined in sRGB gamut, you can see it’s a bit too saturated in the area where the light hits. On the other hand the second image has a much better look with rolloff in highlight in ACEScg gamut especially the green pixels.

OK. A brief conclusion for this short paragraph. If CG Artists use sRGB textures and linear-sRGB working space in their 3D applications, the color in the output images will be limited to the sRGB gamut. In the ACES working space, images will look more realistic with larger gamut and high dynamic range.

ACES for SHOOTING

Right, we introduced the traditional linear workflow and ACES workflow and we pretty much cover the idea for the CG part why we want to use ACES. Let’s recall the workflow graph and take some steps back to the shooting part.

How does ACES affect the production part?

We both know there are a bunch of professional cameras for filming and these manufactures have their own transfer functions(log) and gamuts. In the traditional workflow we need to interpret footage correctly to have accurate calculation and composite with other elements. But here is also the important step that many artists get confused by. Again we take alexa as the example. How do we convert alexa footage into the same colorspace as CG render?

  1. Apply the transfer function to neutralize the footage from logc into linear space.
  2. Convert Alexa Wide Gamut(AWG) into sRGB.

Here is the catch. Lots of artists do not understand the 2nd point therefore they are doing their artistic work in the linear-AWG space instead of linear-sRGB. Well, in many cases, clients don’t care and never get kickbacks and everything is fine. But if you want to know why your CG is fake, it’s better to understand these steps then you will know how to improve it.

In the ACES workflow, your applications will convert the footage whatever its gamut is into ACES and for the rest you just need to do is gamma correction. Awesome. The main purpose is to integrate different footage from different cameras into a big, giant colorspace that you don’t have to worry about having the wrong interpretation.

COMPOSITING in ACES

We have talked about ACES in CG and shooting. It shouldn’t be that complicated now if you already grab the idea above.

As a compositor, you will definitely feel awesome if you get all the elements with the maximum capacity. You know the feeling that a cook has full access to the best ingredients and seasonings. Once you have footage and CG in ACES, you just need to put them into another application in ACES working space and do your compositing. Easy, right? However, we know the idea of ACES, but how do we know exactly what to do in Nuke? I’ll break down the steps in detail in the next post.

Color management is a big topic to me and there are still lots for me to learn, please correct me if there is any inaccurate.