AgX vs Filmic vs ACES vs ARRI

This screen capture of the #Arnold Render View in #Maya 2024.1 with an #OCIO collection for different view transforms demonstrates why #AgX is superior and #ACES is broken.
#Filmic and #ACES can't handle a simple stress test in #Maya using a white sphere and box with a colored bright area light (Intensity: 150, Exposure: 10). The light color is an animation across the #HUE colors and this simple test discovers the issues of #ACES but also shows why #AgX and the #ARRI K1S1 are superior to #ACES.
#ARRI K1S1 is an interesting tone mapping protocol because it's very similar to the #AgX results but with a higher contrast and saturation. This could be helpful for specific cases. But overall the #AgX tone mapping looks more natural and balanced with a perfect grading across the whole image at costs of saturation.
My actual conclusion is that #AgX will desaturate a bit too early. This can be improved/fixed in a final color grading/correction process of course.
#OCIO #Pixelmanager on #GitHub: shorturl.at/acnDV
#Maya scene: bit.ly/3G16B7q

[@piotrus3333_CGI]

Thank you for adding new ACES updates. Regarding ARRI - K1S1 does not give enough justice to their colour. It is outdated as much as ACES 1.2 is. Please consider comparisons using AWG4 and the rest of their REVEAL Color Science. The difference is night and day. And a question: what is the colour value of your white (standard surface I guess?. not a Maya user so I can not check the scene)

[@3dvfxprofessor]

aiStandardSurface with Roughness: 1.0.

[@3dvfxprofessor]

www.arri.com/en/learn-help/learn-help-camera-system/image-science/reveal-color-science

[@piotrus3333_CGI]

@@3dvfxprofessor I ment the value of albedo/base color.

[@3dvfxprofessor]

@@piotrus3333_CGI The standard value of 0.8 (white).

[@piotrus3333_CGI]

@@3dvfxprofessor thank you. all clear.

[@albertolh7513]

Hi! Thanks for the video, how did you apply the AgX view transform?

[@3dvfxprofessor]

WIth Blender. But you can use the AgX OCIO in Maya, NUKE or AFX, too.

[@BoboTheImp]

What is broken with ACES?

[@3dvfxprofessor]

The protocols for view/color transform. Just watch the video. ACES is not able to produce a correct and natural-looking result. There are color shifts and collapsing HUE colors.

[@3dvfxprofessor]

Check my other videos about this topic. Troy Sobotka has a lot of EXR images for testing on his GitHub page. Or just google "Obi-Wan face blue light" and you see good examples of what ACES is messing up. This won't happen with AgX.

[@BoboTheImp]

@@3dvfxprofessor Yes I'm just in the midst of trying to familiarizing myself with these terms. Would you mind clearing something up for me? For instance at 0:51 we see a mostly blue light source hitting the sphere. We are getting a purplish hue on the sphere, and a mostly blue hue on the walls. As my understanding goes, this is from the Abney effect, stating that as more value is introduced to the light, the hue should shift and eventually just appear white at super high values. Two things about this 1. Does this have to do with hue shift? 2. Isn't the Abney effect a natural phenomenon and desirable? I don't see it very much in AgX compared to ACES. I see the shift to losing saturation but not much hue shifting. Is the ACES form of this too pronounced maybe? I'm not sure I can detect hue collapse yet, but I'm looking at your videos and others.

[@3dvfxprofessor]

Hi@@BoboTheImp The HUE shift has nothing to do with the Abney effect. The Abney effect describes the perceived color shift by the stimulus that occurs when white light is added to the light of a monochromatic light source. What you see in the video is a result of the ACES color transform protocol which does it wrong. The Abney effect would occur when you have a blue color being desaturated towards white. The stimulus lets your eyes and brain see purple colors in between. When you use a colored pure blue light source shining onto a white background the background looks blueish. If you raise the light intensity the background will get brighter but not bluer than blue. That's what AgX is doing correctly. And this is the problem with the ACES View Transform. ACES tries to keep the purity of colors and doesn't desaturate the colors toward white. This ends up in a HUE collapse because there is no higher blue than the blue of the HUE spectrum. The color shift from blue to purple is a specific error in the protocol. The newer ACES 2.0 is doing something similar to AgX now in this scene. But it still looks oversaturated. Now folks would say that this is fine now but when you do further tests with a a different setup (check my latest videos) you will still get bad results while AgX is doing it perfectly. Troy Sobotka has published an archive of EXR images with extreme situations on GitHub for testing. The only protocol which is able to deal with all of them is AgX. Every other protocol fails. Especially ACES.

[@3dvfxprofessor]

@@BoboTheImp, you can see the HUE collapse very clearly here: kzbin.info/www/bejne/bV7dd4qoqrN9bbc