Physically inacurrate light distribution in Octane Tweet

[SSmolak]

Please see ambient light scenario compared to real photo. Octane completely ignore indirect light bounce to polygons that are not directly exposed to ambient sky. I have the same colors and the same ground brightness as in real photo.

There is no light bounce to ceiling from floor and adjected walls.

It even partially ignore reflections :

Please don't tell me that this is physically accurate.

To make the same brightness in Octane there is need to increase ground brightness but in real scene ground isn't that bright - there is more grass around than pavement. Also there is need to fully increase diffuse roughness that works only for Diffuse material ( problem described here : viewtopic.php?f=23&t=80314&p=414925#p414925 )

Increasing sky brightness do nothing for polygons oriented directly to ground. There is no light bounced from walls to them - only from ground that is very strange...

more references : https://www.shutterstock.com/pl/search/ ... chitecture

[SSmolak]

I found solution :

The problem was Octane Hosek&Wilkie implementation but not only.

1. Hosek&Wilkie implementation is totally not suitable for creating ambient, cloudy lighting because even without sunlight increasing turbidity create very large, bright halo on the sky. This halo cause shadowing even if it is placed near sky zenith - it create shadow under balconies. The best is to use cloudy HDR maps or old Octane daylight model where we can avoid this turbidity halo and make more uniform lighting.

2. Using diffuse, color, albedo ( no matter what it's called ) on walls should be avoid or with little percentage only. For little irregularity, bumps the best is to use normal maps. Octane do it right for simulate little bumps, cracks etc with natural darkening without the need of use albedo maps. AO maps should be not used at all - they are creating fake shadow effect that darked whole surface. Use albedo only for color changes or dirts.

3. There is need to use roughness on diffusion too. Octane use it only in Diffuse Material. For all other materials roughness is used only for specularity, reflections... To make more soft light distribution on corners there is need to use Diffuse Material with roughness like 30-75% and if you want specularity - use Specular Layer.

After that we have even more bright polygons in corners or under balconies than in reality where in typical situation main wall has 45% RGB brightness, polygons under 30% and ground average 40%.

[SSmolak]

I decided to wake up this thread because after year of making architectural viz's, testing, comparing renderings with other engines and self investigation, photos I'm 100% sure that Octane has serious issue in the terms of light distribution in not only dark areas but especially with polygons that are not directly exposed to direct light !

This is why most of the people use Corona and VRay for such type of renderings.

Within the time I saw people that asked the same, some days ago for example : the same interior scene in Corona and Octane : https://www.facebook.com/groups/OctaneR ... 3274330068

Problem is with :

1. When in some areas light power is decreased gamma fallow for materials in Octane is decreased too much - it create too much darkness in dark area. Also color of material distribution to other material is decreased too. It create somewhat flat areas especially in interiors where light from brown wood floor should be bounced more.

2. Octane doesn't take in mind that light bounced from bright wall to another bright wall should increase its power to make wall that it is oriented 90 degrees to them should receive more light power that from outside environment - like these walls above balconies that are not directly exposed to direct and indirect light from sky or Sun.

For making exterior ambient light like cloudy sky sometimes best is to use Direct Lighting with Ambient Occlusion mode especially in dense, tight areas ! This is nonsense :

Example of dark area but should be more bright because of 80% bright ground where light should be bounced to up :

Underexposed parts :

O1.jpg (24.79 KiB) Viewed 1275 times

O2.jpg (18.63 KiB) Viewed 1275 times

Real photo, ambient, cloudy sky. Balconies obstructed by trees. Pure white even on walls under - impossible to make in Octane :

[elsksa]

https://render.otoy.com/forum/viewtopic.php?f=9&t=79300&p=410178#p410178

[SSmolak]

https://render.otoy.com/forum/viewtopic.php?f=9&t=79300&p=410178#p410178

Yes this works fine like simple scenes but unfortunately not for more complex areas. Of course it can be done in post - but why ? The question is why Octane has issues in this area but other renderer not ? I thinking about mixing AO with beauty pass in AOV masking with walls but this is still workaround. This should be really make thinking for devs - especially these badly dark areas which I sent.

Octane has Diffuse level but more that 6 do nothing, the same for GI - more than 100 do nothing too. But max level is set to insane values - strange...

There is no problem if direct light like Sunlight is bounced somewhere to the problematic polygons, but if we have only indirect light - the problems are coming.

Look on this lightness degradation from left to right where column on the same position is bright, after on the right there is still dark - only on polygons towards ground. Ground is pure bright but Octane indirect light can't transfer as much energy to up.

[elsksa]

Do you have any of the aforementioned renderer comparison images at your disposal?
Same data set (scene), color pipeline, shading, render settings (equivalently matched) with screen grabs of settings for "peer-checking".

Nothing will be efficiently convincing as it is objectively not an issue.

None of the renderers you've mentioned have a particularly special "GI rendition" or are known for it.
It is beyond a doubt possible in Octane, as shown in my previous illustrated reply.

Indubitably, there isn't any major concern regarding Octane's GI.
Increased shading and lighting values, under acceptable limits, is often what's being neglected. This applies to cloud lighting as well.

[SSmolak]

Do you have any of the aforementioned renderer comparison images at your disposal?

I will share such scene with comparison to Corona soon.

Meanwhile, thanks to great Octane AOV, mixing by 20-35% Ambient Occlusion with beauty pass in screen mode and masking it by material mask, channel map range to get only dark areas and white balance adjusting because Occlusion is white can do great job for resolve my problem. I'm very happy with this solution. AOV is very powerful. This is still workaround but very good configurable.

AO_AOV.zip

(540.66 KiB) Downloaded 20 times

It works great with irradiance channel too

irradiance.zip

(316.78 KiB) Downloaded 21 times

[SSmolak]

Corona :

Octane :

As you see they are totally different in dark areas. Even Octane is darker at all using the same sky color !

Ha ! It looks that Octane needs 2.2 gamma ( sRGB response ) to looks the same as Corona :

Probably VRay use the same gamma as Corona. That's very interesting, why Octane use like double gamma linearity compared to other and it looks overall darker by default and needs more color saturation to compensate this.

[elsksa]

Same data set (scene), color pipeline, shading, render settings (equivalently matched) with screen grabs of settings for "peer-checking".

Indeed.

With a superior rendering transform than default obsolete color management, a good lighting, shading
and appropriate kernel settings, there is absolutely no issue at getting a "high light bounce look". The same goes to other renderers as it's pretty much the same story for all of them.

[SSmolak]

But when we compare direct light distribution we have proof how different these engines works. Increasing gamma in Octane washed too much of whole scene and HDR map that looks different :

Default 1.0 Octane gamma makes it way to dark compared to Corona :

So as you see Octane light distribution using the same HDR is much less powerful than in Corona.

Also look how different and lack of light is in this place in Octane :