Hosek&wilkie - ACES color compensation

Generic forum to discuss Octane Render, post ideas and suggest improvements.
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Post Reply
User avatar
SSmolak
Licensed Customer
Posts: 1157
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Contact:

Every of us know that ACES can be bad in the terms of color shifting. Mainly blue color that is shifted to cyan. This color shifting is mainly visible using Hosek&Wilkie sky on the reflections like window glass. Strange is that it is not visible on Sky itself.

I think that H&W Sky model needs to have switch to compensate this color shifting. Probably this sky model shouldn't be used with ACES. For example, ACES P30-D60 looks great without blue color shifting but it has less saturation and contrast.

Probably there is something wrong with Hosek&Wilkie blue sky color that is not suitable to regular ACES. It should be little moved to the more temperature in Kelvin scale so it will be more blue not cyan like.

sRGB :
sRGB_.jpg
ACES :
ACES_.jpg
Attachments
blue_shift_aces_srgb.zip
(353.99 KiB) Downloaded 78 times
Architectural Visualizations http://www.archviz-4d.studio
elsksa
Licensed Customer
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

Hi,

Nothing much to do on the Octane side (unless an aspect of it has been overlooked).
ACES is flawed (period) and the rest would rather be a post matter (EXR output mandatory).

Worth to mention that the Octane Camera Imager features a "white point" option.
User avatar
SSmolak
Licensed Customer
Posts: 1157
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Contact:

Right, but this cyan cast on dark blue reflected areas is from Hosek&Wilkie sky model. Other colors looks fine as should be.
Architectural Visualizations http://www.archviz-4d.studio
elsksa
Licensed Customer
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

ACES does not output the same imagery as with the default sRGB color management setting. That is to be expected.
Some other color encoding pipelines do produce different results as well.
Nothing is set to stone, none of them are impartial (objective, neutral).

This can be handled in post, with the necessary know-how, which I understand, not everyone has.
User avatar
SSmolak
Licensed Customer
Posts: 1157
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Contact:

elsksa wrote:ACES does not output the same imagery as with the default sRGB color management setting. That is to be expected..
And this is why we have OCIO in Octane LiveViewer - to make look of what we should expect in final - maybe not exactly 100% perfect but near. If not, we can always render as ACES without previewing it in LV right ? Artists wants to see in previewer as much as possible final result before doing tonemaping in DaVinci, Fusion or AfterEffect. This is simple to do.

And please not say that artists using 3d softwares from 15 years in the terms of modelling, lighting, surfacing and processing them in post-softwares are soo stupid to using ACES.
Architectural Visualizations http://www.archviz-4d.studio
User avatar
SSmolak
Licensed Customer
Posts: 1157
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Contact:

ok please close thread :)
This color cast can be reduced or changed using Hosek&Wilkie ground color. It works better than changing whole image colors by Camera White Point.

Also I compared H&W clear sky to one of the best clear sky HDRI maps and I think that HDRI has better colors than H&W using ACES by default.
Architectural Visualizations http://www.archviz-4d.studio
elsksa
Licensed Customer
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

SSmolak wrote: Also I compared H&W clear sky to one of the best clear sky HDRI maps and I think that HDRI has better colors than H&W using ACES by default.
There is no such a thing as "better colors". This is purely subjective. The concept of color only exist through our own eyes as "the observer".

A sky model is only a simulation (more or less) of the reality.
An HDRI is a subjective (biased) digital camera and formed imagery result of what was photographed at a given location and time.

"Best" here is "your own"! Last but not least, anything ACES is prone to break. That's inevitable.
User avatar
SSmolak
Licensed Customer
Posts: 1157
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Contact:

What we see on digital monitor even best calibrated is not the same that our eye see in real world. I spent two months to make photos of vegetation and buildings in different light conditions. Compared this with my own eye after that and color picker, simulated the same in Octane I can say that ACES can do that - but with using proper camera exposure.

Human eye has very wide exposure. Camera exposure system works different. It has point. Octane exposure is another different thing. To use ACES it should be set different camera exposure for every shot.
Architectural Visualizations http://www.archviz-4d.studio
elsksa
Licensed Customer
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

SSmolak wrote:What we see on digital monitor even best calibrated is not the same that our eye see in real world.
Exactly, yet, unsurprisingly! A monitor **emits light**, what we see is **reflected & transmitted light**.
SSmolak wrote: I spent two months to make photos of vegetation and buildings in different light conditions. Compared this with my own eye after that and color picker, simulated the same in Octane I can say that ACES can do that - but with using proper camera exposure.
You are missing the point.
1. ACES is, from the ground up, from its core and design, a broken C.E.S..
Nobody can't deny these actual facts. That's not even up to a debate or argumentation. Whoever is free to use it but can not claim anything pragmatically efficacious.
2. To sum it up: two months of digital photography, taken with a specific sensor, raw data formed into a "viewable imagery" using camera manufacturer basis and a software (likely PS or LR). Do you see where I am heading?
SSmolak wrote: Human eye has very wide exposure. Camera exposure system works different. It has point. Octane exposure is another different thing. To use ACES it should be set different camera exposure for every shot.
"Dynamic range"*, not exposure. The rest of the phrase is equivocal.
•The human visual system is paired with the brain. Think of it as lens + sensor + the processing to produce a viewable result of what is seen - all of which is automatically and constantly adjusting (including the many visual illusions!), such as chromatic adaption which anyone can experience with a couple of lights ranging from a "warm" to "cold" (likely color correlated LEDs these days) kelvin temperature. Even more with strong "RGB" light sources. The same goes for the iris, which automatically adjust itself depending on the situation, and so on.
• A camera (film or digital) is a naïve tool controlled by a user, featuring a light-sensitive surface. Its exposing to light is always user-controlled.
• Octane and any other renderer are computer-simulated imagery software with a particularly major distinction from scanned film and digital cameras: floating point in-renderer and at encoding level as opposed to integer encoded digital imagery.

It's not up to the developers to build a specific subjective look for every user's taste. That's an infinite amount of work and not feasible. Not to mention that it would require a dedicated department (R&D + financial and human resources) to develop a viable solution, which only a handful of companies are close to, these days. I let you imagine what it takes.
The benefits of EXR encoded file is to offer to the knowledgeable and skilled user, the choice of the color pipeline. Putting aside any in-renderer bugs, issues and whatnot.

Even if developers implement some built-in "looks" (apart from the already deprecated ones present), it will never be as solid and customizable as to what can be done in post or through OCIO (which one of its goals is to bring the post stage as a preview during rendering, in such "CGI" context), not without what was previously mentioned (i.e. the necessary resources).

That's the tradeoff of producing "images" via computers instead of photographing on film or a digital camera (where for the latter, camera manufacturers do a partial work on the digital imaging development). Everything is done from scratch with a renderer. From the bits and line of codes to the viewable results of a simulated 3D scene.
User avatar
SSmolak
Licensed Customer
Posts: 1157
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Contact:

Many great info here. Thank you.

I still think that exposure is the most important thing to work with ACES in Octane live viewer if we want to see near the same as we must do in post.

Some time ago I asked about auto-exposure feature here : viewtopic.php?f=9&t=80367&p=416296&hili ... re#p416296

Meanwhile I did script which automate exposure in relation to Sun position vs Camera.

Look on info :
info.jpg
info.jpg (20.98 KiB) Viewed 1598 times

Here is ACES without auto-exposure :
Aces_locked_exposure.zip
(3.63 MiB) Downloaded 69 times
ACES with auto-exposure :
Aces_auto_exposure.zip
(4 MiB) Downloaded 71 times
comparison :
locked_vs_auto.zip
(464.02 KiB) Downloaded 67 times
Architectural Visualizations http://www.archviz-4d.studio
Post Reply

Return to “General Discussion”