Physically inacurrate light distribution in Octane

Forums: Physically inacurrate light distribution in Octane
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

Re: Physically inacurrate light distribution in Octane

Postby elsksa » Tue Jan 02, 2024 9:00 pm

elsksa Tue Jan 02, 2024 9:00 pm
""Gamma"" (transfer function) is utterly irrelevant here. In other words, it is here a technical parameter that cannot be or influence as a shading or GI factor.

In fact, 1.0 ""Gamma"" (aka linear transfer-function) is one way to ensure incorrect output/display (of the formed image) encoding as it's a non-conformed state. For comparisons, not that I endorse them, it should be an identical color pipeline or at least, transfer function. Worth mentioning that Octane is a spectral renderer (on this page, it is briefly explained what it means).

I invite you to thoroughly consult this guide on Color Management as well.

Regarding the results, it is impossible to judge an image without scene kernel settings shading and lighting information. A "brighter white" diffuse in Octane, with proper settings will result identically to the "reference".
elsksa
Licensed Customer
Licensed Customer
 
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

Re: Physically inacurrate light distribution in Octane

Postby elsksa » Tue Jan 02, 2024 9:47 pm

elsksa Tue Jan 02, 2024 9:47 pm
A more suitable approach for comparisons is the use of legally sharable assets such as the popular Sponza, considering the intention of this thread.

Nothing scientific in the following examples, as that's pointless.

Sunny Nishita
Image

Cloudy HDRI (Camera Imager "White Point" not adapted)
Image
Attachments
sponza-cloudy-octane.jpg
sponza-scene-octane.jpg
elsksa
Licensed Customer
Licensed Customer
 
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

Re: Physically inacurrate light distribution in Octane

Postby SSmolak » Wed Jan 03, 2024 2:14 am

SSmolak Wed Jan 03, 2024 2:14 am
Thank you, your technical knowledge is very helpful for understanding how things works in renderers but unfortunately this not help for creating the same image look that offer another renderer because developers using the same technical data implemented them little different.

After another day of investigation and testing, finally I found what is going on !

I know now, why people interested in creating good looking exterior images choose Corona or Vray. Not because Octane is worse. Because these renderers oriented for exteriors have great built-in calibrated settings where you can create simple materials, using built-in light system, pressing render and excellent, great pleasure for eye image is done. In Octane to achieve the same you must tweaking. People are lazy in nature.

From left to right - Octane, Corona. This time without cheating with gamma or ambient occlusion. The same camera exposure.

HDR Lighting :
Octane_Corone_Hosek.png


Hosek & Wilkie lighting :
Octane_Corona_HDR.png


You can get color picker, I prefer MS PowerToys which has on-screen HSV picker https://github.com/microsoft/PowerToys/ ... ag/v0.76.2 and you can see that every the same part of images have the same color, saturation and brightness level with error max 5%.

They looks EXACTLY the same. So where is problem ? Why 80% of the architectural visualization market use Corona, VRay not Octane ? Because Octane has badly implemented material brightness calibration and bad default values for SkyLight !

1. Main problem in Octane are materials. They are strange compared to Corona because Octane color brightness on 80% are the same that Corona use on 50%. When you want more photons to be pushed out and lighting darker areas you must use 15-35% more material brightness for that here. The same situation is for color saturation. Octane needs 5-15% more saturated color for create the same color expansion from one material to another. Especially visible in interiors.

Material brightness in Corona from 50% to 100% are the same in Octane using 85-100% values. Something wrong is with calibration here. We have much more possibility to set slider within material darkness than in its brightness.

2. Octane Hosek @ Wilkie looks little different in Corona. The most important thing is that Octane Sun Power is way too strong by default. It create too dark shadow areas in relation to Sky power. To make it the same as in Corona it should be set to 0.4 if Sky Power is set to 1. After that we have more brightness in shadow areas with not too much dark shadow.

This solution works fine for simple colored materials but not for textured ones. Every color texture needs to have much more power to achieve the same bright expansion in GI by using additional color correction node or by power in image node - still needs to tweak.

I will share these scenes for both Octane and Corona so you can play with it within day or two.
Architectural Visualizations http://www.archviz-4d.studio
User avatar
SSmolak
Licensed Customer
Licensed Customer
 
Posts: 1095
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland

Re: Physically inacurrate light distribution in Octane

Postby elsksa » Wed Jan 03, 2024 12:56 pm

elsksa Wed Jan 03, 2024 12:56 pm
SSmolak wrote:but unfortunately this not help for creating the same image look that offer another renderer because developers using the same technical data implemented them little different.

Precisely.
I believe Corona Renderer (still) has its own custom internal color pipeline, merely a wider Gamut (rendering/working color space).

Matter of fact, not even the Autodesk Standard surface is implemented identically in some renderers. Although the dissimilarities are too subtle to be qualified as evident.

SSmolak wrote:Because these renderers oriented for exteriors have great built-in calibrated settings where you can create simple materials, using built-in light system, pressing render and excellent, great pleasure for eye image is done. In Octane to achieve the same you must tweaking. People are lazy in nature.

If people would thoroughly read the links Ive shared numerous times, they would have noticed the recommendations echoing your "findings".

"Calibration" would be a misnomer subject to avoidance.

SSmolak wrote:1. Main problem in Octane are materials. They are strange compared to Corona because Octane color brightness on 80% are the same that Corona use on 50%. When you want more photons to be pushed out and lighting darker areas you must use 15-35% more material brightness for that here. The same situation is for color saturation. Octane needs 5-15% more saturated color for create the same color expansion from one material to another. Especially visible in interiors.

Material brightness in Corona from 50% to 100% are the same in Octane using 85-100% values. Something wrong is with calibration here. We have much more possibility to set slider within material darkness than in its brightness.

This isn't a scientific or statistical study. Nobody can be spitting percentage numbers without an in-depth contextualized researched.
It's not about science or physical accuracy, renderers are not scientific tools but an over simplification of the infinitely intricate reality. It takes some advanced mathematics to build a renderer and great efforts to design it in a user-friendly way (non-scientific and practical).

SSmolak wrote:2. Octane Hosek @ Wilkie looks little different in Corona. The most important thing is that Octane Sun Power is way too strong by default. It create too dark shadow areas in relation to Sky power. To make it the same as in Corona it should be set to 0.4 if Sky Power is set to 1. After that we have more brightness in shadow areas with not too much dark shadow.

There is no ground truth default. Sometimes one (sky or sun intensity) needs to out-power the other to achieve a certain look. As long as one or the other or both are higher values than default.
Needless to mention that the kernel settings have to be appropriately set as previously mentioned and showed on the aforementioned links.

SSmolak wrote:I will share these scenes for both Octane and Corona so you can play with it within day or two.

Appreciate the intention, but that won't be necessary as I won't be in a capacity to download any.

To summerize,
• comparisons aren't fairly feasible due to the many diverging factors
• renderers are not equally developed and designed yet closely similar modern technologies
• implementation differs between renderers, although subtle in some cases
• renderers aren't scientific tools but image-making tools
• any modern renderer is capable to achieve "high bounce/GI look" with appropriate shading, lighting and settings values
• exposure principles from photography also applies to rendering, an interior sunny day is expected to illuminate an interior but be relatively blown out at the windows, requiring a Camera Imager exposure adjustement and vice versa.
• a proper and correct color pipeline is mandatory
elsksa
Licensed Customer
Licensed Customer
 
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

Re: Physically inacurrate light distribution in Octane

Postby SSmolak » Wed Jan 03, 2024 5:09 pm

SSmolak Wed Jan 03, 2024 5:09 pm
To be clear - my intention was to report that Corona can produce great looking images using default settings. It produce quality which is standard in industry. In Octane to achieve the same you must tweak default settings and texture power within materials.

As I know many people think that this is related to limitation in render engine - not settings. I proofed that they are wrong but that doesn't mean everything is fine because for many people finding what is needed to tweak to achieve intended result can be too much to stay with Octane.
Architectural Visualizations http://www.archviz-4d.studio
User avatar
SSmolak
Licensed Customer
Licensed Customer
 
Posts: 1095
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland

Re: Physically inacurrate light distribution in Octane

Postby elsksa » Thu Jan 04, 2024 2:13 pm

elsksa Thu Jan 04, 2024 2:13 pm
SSmolak wrote:To be clear - my intention was to report that Corona can produce great looking images using default settings

A "biased" default presumably more suited to their user base.
Such substantial change wouldn't be sensical without statistics and user-base polls to back it up.

SSmolak wrote:It produce quality which is standard in industry. In Octane to achieve the same you must tweak default settings and texture power within materials.

The same goes to industry leading CPU renderers used by small to large VFX studios and digital agencies.

SSmolak wrote:As I know many people think that this is related to limitation in render engine - not settings. I proofed that they are wrong but that doesn't mean everything is fine because for many people finding what is needed to tweak to achieve intended result can be too much to stay with Octane.

People's ignorance comes from their lack of education on a given subject. Unlike, let's say the healthcare sector, nobody is forced to learn computer graphics and it shows. This isn't at the responsibility of developers.

A plethora of gorgeous images created with Octane would suffice to convince that the tools is capable in good (skilled and knowledgeable) hands. As for any tool. Or as little as the technical specifications of the renderer.

That said, there are some default parameters that should be changed by default. Off the top of my head:
• adaptive sampling enabled by default
• GGX Energy Conserving as default "BRDF model"
• Path Tracing Kernel with maxed out Caustics Blur (particularly since the addition of the Photon Tracing Kernel)
• Area lights with a pure black diffuse "color"
elsksa
Licensed Customer
Licensed Customer
 
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

Re: Physically inacurrate light distribution in Octane

Postby SSmolak » Fri Jan 05, 2024 12:50 am

SSmolak Fri Jan 05, 2024 12:50 am
elsksa wrote:Such substantial change wouldn't be sensical without statistics and user-base polls to back it up.


2022 Architectural Visualization Rendering Engine Survey Results : https://www.cgarchitect.com/features/ar ... ey-results

Octane can produce the same or even better quality than Corona or VRay but most of the people doesn't know about or they didn't get good result by using default settings.
Or maybe they scare about VRam limitation, Nvidia card cost relation to CPU - I don't know but 2,45% market usage is hmmm bad.

Interesting is that default Max Scanline has more usage than Octane. It is very bad renderer. Even Redshift has more. Twinmotion, Lumion ? What the hell . I think that people use renderers because other use and they are more popular.
Architectural Visualizations http://www.archviz-4d.studio
User avatar
SSmolak
Licensed Customer
Licensed Customer
 
Posts: 1095
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland

Re: Physically inacurrate light distribution in Octane

Postby elsksa » Fri Jan 05, 2024 2:06 pm

elsksa Fri Jan 05, 2024 2:06 pm
SSmolak wrote:
elsksa wrote:Such substantial change wouldn't be sensical without statistics and user-base polls to back it up.


2022 Architectural Visualization Rendering Engine Survey Results : https://www.cgarchitect.com/features/ar ... ey-results


That's too light to be qualified as a reliable data.
Most surveys aren't trustworthy, objective or particularly informative.
My interest in renderers comparisons is pretty much non existent. I am the type to not care about industry trends and to believe that tools are just tools, whatever floats the boat and does the job. As it should be.

Since we're here, one of the reasons why Octane isn't as used as X or Y renderer in that niche branch of CGI goes way beyond its development.
A lot of online assets made in more or less the past decade, are intended for these popular renderers.
The limitations of GPUs themselves plays a substantial role as well.
CPU based workstations are more forgiving on lack of optimisations and technically allows systems to run on hundreds to thousands of RAM. Granted CAD data involvement in the rendering pipeline, scattering of non-optimized assets and so on, it becomes evident.

Nonetheless, GPU limitations are shrinking, from both a physical/engineering perspective and software development improvements. Apple broke some barriers already with laptops (and their "Studio" lines) holding more than a hundred of shared RAM. The (near) future is certainly promising. Time will tell.

My short 2 cents.
elsksa
Licensed Customer
Licensed Customer
 
Posts: 784
Joined: Sat Jul 24, 2021 1:06 am

Re: Physically inacurrate light distribution in Octane

Postby SSmolak » Wed Jan 31, 2024 7:21 am

SSmolak Wed Jan 31, 2024 7:21 am
elsksa wrote:My short 2 cents.


Thank you for explanation. Your knowledge is always welcome.

My last two cents : Octane can produce insane images and without issues that I described before but it needs much more tricks to achieve the same. Mainly because of different materials treatment. Problem is not lighting, engine but materials.

oct.png


oct2.png
Architectural Visualizations http://www.archviz-4d.studio
User avatar
SSmolak
Licensed Customer
Licensed Customer
 
Posts: 1095
Joined: Sat Feb 07, 2015 5:41 pm
Location: Poland
Previous

Return to General Discussion


Who is online

Users browsing this forum: No registered users and 39 guests

Sat Apr 27, 2024 10:17 am [ UTC ]