Isotropix Forums

Clarisse 4.0 light shadow bug? SOLVED

General Discussion about Isotropix and CG related topics

Clarisse 4.0 light shadow bug? SOLVED

Unread postby atnreg » Sat Feb 02, 2019 6:42 pm

Hi!

I just noticed strange behaviour which seems bug to me but as I may be wrong, I decided to write here first :)

clarisse_4_possible_bug.png
clarisse_4_possible_bug.png (39.03 KiB) Viewed 1412 times

clarisse_4_lightbug.project
(46.25 KiB) Downloaded 30 times


Please see the attached project, there is only one distant light, one sphere light and one polygrid, nothing else.
I cannot understand why the sphere light casts shadow even when it is Unseen by camera. I think lights should never cast shadows and that does not happen in Clarisse 3.6. In 4.0 that happens even with Legacy mode :o
Shadow appears when the Radius is very small but when radius is increased, the shadow appears and it is impossible to get rid of it.

Am I doing something wrong or is this bug? :)

Thank you!

Antti
Last edited by atnreg on Wed Mar 06, 2019 6:28 pm, edited 1 time in total.
Antti
AMD Ryzen Threadripper 2990wx (32c/64t),64GB RAM,NVIDIA GTX 1080ti,Win10
Clarisse 4.0SP4c,Houdini,Blender2.80,Fusion360,Onyx,ZBrush,SubstPnt...
Started: Clarisse 2016/10 (Py 2017/01), Python 2016/11
No business, just fun :)
atnreg
 
Posts: 511
Joined: Mon Sep 19, 2016 4:20 pm
Location: Helsinki, Finland

Re: Clarisse 4.0 light shadow bug?

Unread postby vandam » Sat Feb 02, 2019 7:34 pm

Hmmm, it is different than in other engines (Arnold lights do not cast shadows by themselves, in Vray they do but once you set them to invisible they stop), but whether it's a bug or intended behaviour - I cannot say. I'm leaning towards a bug - since there is no light-centric light linking in Clarisse you can't exclude one light from incoming illumination of another light. So there would be no way of getting rid of this shadow.
Cheers
vandam
 
Posts: 148
Joined: Wed Jun 21, 2017 9:22 pm

Re: Clarisse 4.0 light shadow bug?

Unread postby atnreg » Sat Feb 02, 2019 9:13 pm

vandam wrote:Hmmm, it is different than in other engines (Arnold lights do not cast shadows by themselves, in Vray they do but once you set them to invisible they stop), but whether it's a bug or intended behaviour - I cannot say. I'm leaning towards a bug - since there is no light-centric light linking in Clarisse you can't exclude one light from incoming illumination of another light. So there would be no way of getting rid of this shadow.
Cheers


Exactly, it would be ok if it was possible to avoid it too but as normally (ever?) we don't want to see shadows from light sources themselves (unless there is actual geometry meant to be visible but that is usually separate geometry from the light source) I think this is not intentional. And also because 4.0 behaves differently also in legacy area projection mode (spot) so it is not only the new light system that behaves like this (actually the legacy mode also probably uses the new system but still using it should give same result as before) :)

Antti
Antti
AMD Ryzen Threadripper 2990wx (32c/64t),64GB RAM,NVIDIA GTX 1080ti,Win10
Clarisse 4.0SP4c,Houdini,Blender2.80,Fusion360,Onyx,ZBrush,SubstPnt...
Started: Clarisse 2016/10 (Py 2017/01), Python 2016/11
No business, just fun :)
atnreg
 
Posts: 511
Joined: Mon Sep 19, 2016 4:20 pm
Location: Helsinki, Finland

Re: Clarisse 4.0 light shadow bug?

Unread postby sam » Sun Feb 03, 2019 8:22 am

So technically this is not a bug. Likes are physical geometries emitting lights. They are not "explicitly" casting shadows but they are occluding themselves (which lead to a shadow because of lack of illumination in the occluded area). This is basically an eclipse you are seeing.

The thing is we went for the fully energy conservative road. In Clarisse, lights are always transmitted, reflected etc... If you don't do light occlusion then reflections becomes additive for example. If you only do light occlusion for let stay reflection and not diffuse then the energy conservation is wrong thus leading to noise and artifacts.

However, in the same way geometries can be prevented to cast shadows (and cast self shadows in 4.0), we could add this as a feature I guess.
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 1434
Joined: Fri Jan 25, 2013 11:33 pm

Re: Clarisse 4.0 light shadow bug?

Unread postby vandam » Sun Feb 03, 2019 10:42 am

Remembering all the stuff you had said about the future of Clarisee I was actually thinking that this was intended to work this way. I, for one, like this approach - the more PBR the better. It might be inconvienent now in some situations but once you introduce the new emissive material I will not care.
Cheers
vandam
 
Posts: 148
Joined: Wed Jun 21, 2017 9:22 pm

Re: Clarisse 4.0 light shadow bug?

Unread postby atnreg » Sun Feb 03, 2019 11:26 am

Ok so it is technically not a bug but that behaviour makes it impossible to make effect lights which is a common way to highlight some objects or areas. Of course physically and optically a light source always casts a shadow if the overall light is brighter but as we already have option to prevent geometry from casting shadows, that must be possible for lights as well.

And as there is already option to say Unseen by camera for the light, that should of course handle this so that the light itself indeed is not seen by the camera :)

Problem is even more notable if you want to use spot light behind an plane light as effect. Now the spot light does not light anything as the plane light blocks its light. That cannot be correct when Unseen by camera is checked :O

So please at very least add option to make light itself REALLY invisible or anyway somehow make it possible to use lights as in 3D world effects like all other software I know does (including Clarisse 3.6) :)

Thank you!

Antti
Antti
AMD Ryzen Threadripper 2990wx (32c/64t),64GB RAM,NVIDIA GTX 1080ti,Win10
Clarisse 4.0SP4c,Houdini,Blender2.80,Fusion360,Onyx,ZBrush,SubstPnt...
Started: Clarisse 2016/10 (Py 2017/01), Python 2016/11
No business, just fun :)
atnreg
 
Posts: 511
Joined: Mon Sep 19, 2016 4:20 pm
Location: Helsinki, Finland

Re: Clarisse 4.0 light shadow bug?

Unread postby vandam » Sun Feb 03, 2019 12:13 pm

Just out of curiosity - why would you place a spotlight behind area light?
Also - keep in mind that primary visibility (or visibility to the camera) does not always equal visibility to the rays.
I'm doing some freelancing stuff at the moment so I can't really get my hands on 4.0. I've been lighting my stuff using area lights for quite some time now, most of the time lighting is slightly faked to achieve desired results, so I'm curious how 4.0 would affect this kind of workflow.
Some time ago I did something like this:
https://www.artstation.com/artwork/zLkAq
I guess it would not be possible with one light blocking the other. On the other hand - before I started working in a studio - I would never light stuff this way. I felt like it was cheating, or that it is not how light would behave in real world. I guess with 4.0 I'll go back to the roots, no more faking stuff. I'm anxious to try it out. Since interiors are like the worst case scenario - if it works fine without faking out too much - I will not care for this.
vandam
 
Posts: 148
Joined: Wed Jun 21, 2017 9:22 pm

Re: Clarisse 4.0 light shadow bug?

Unread postby K_A_B_Art » Sun Feb 03, 2019 12:33 pm

Hi!

Ouch!
For me this is a real issue, since I’m used to build my lighting with multiple ‘layers’. Now it seems that with Clarisse 4.0 most of my lights will be left in the shadows (witch I do not understand) of the nearer lights.

Example pictures attached. Same scene rendered with Clarisse 3.6 and Clarisse 4.0. With Clarisse 4.0 the area light somehow blocks all the light emitted by spot light.

Scene: Blue area light, purple spot light, a sphere and a polygrid:
Cl_3.6_3D.JPG
3D view Clarisse 3.6


Rendered with Clarisse 3.6:
Cl_3.6_image.JPG
Image view Clarisse 3.6


Rendered with Clarisse 4.0:
Cl_4.0_image.JPG
Image view Clarisse 4.0. Red spot is blocked by area light.


Please do consider fixing this.

Cheers!
Kim
K_A_B_Art
 
Posts: 14
Joined: Mon Jan 02, 2017 6:58 pm

Re: Clarisse 4.0 light shadow bug?

Unread postby atnreg » Sun Feb 03, 2019 1:11 pm

The 4.0 is acting like simulator but 80% of the users want to use any 3D software in artistic way i.e. making even impossible scenes to achieve some effect. It is great that 4.0 has this simulation behaviour and geometry acting as lights soon too but it must be an option, not the only way to do things.
As the above posts show, many existing, even basic scenes won't work at all in 4.0 anymore because of this.

So please make an option to get real invisible lights in 4.0 :)

Antti
Antti
AMD Ryzen Threadripper 2990wx (32c/64t),64GB RAM,NVIDIA GTX 1080ti,Win10
Clarisse 4.0SP4c,Houdini,Blender2.80,Fusion360,Onyx,ZBrush,SubstPnt...
Started: Clarisse 2016/10 (Py 2017/01), Python 2016/11
No business, just fun :)
atnreg
 
Posts: 511
Joined: Mon Sep 19, 2016 4:20 pm
Location: Helsinki, Finland

Re: Clarisse 4.0 light shadow bug?

Unread postby mdkai » Mon Feb 04, 2019 1:39 am

I had to think about this as well for quite some time if i would consider this a problem or not, and i understand the explanation Sam was giving. Dealing with exceptions is often a fine line between doing good and evil.

But in this case it is absolutly crucial if lights occlude other lights, as some combinations are simply impossible to light without revealing the lights itself.. i.e.: a single directional light plus any finite light type with radius will reveal its presence in the scene without the artist having any control about it.
lightshapes.PNG
revealing light shapes by directional light


Unless there is a way to overcome it, i would consider this to be really important to find a better solution as there is now way of preventing the lightshape occlusion for now ..

Sam wrote: If you don't do light occlusion then reflections becomes additive for example.


But even if physically not plausible to light twice (or additive) i think i would prefer that over the current state, which can lead to strong visual artifacts, that can hardly be fixed, no ?

Given that the new light system took it's toll and will offer a much more revolutinary approach, i am very confident this will be resolved as well.

Cheers
Kai
Join us at the Discord https://discord.gg/G5cJj9A
mdkai
 
Posts: 310
Joined: Tue Oct 07, 2014 7:24 pm

Next

Return to General Discussion