Blender Indirect light test and Yafaray comparison

![][IMG]](][IMG] is a scene I made for a project which is light only with a sun lamp and a little emit on the background sky texture. Here are the 3 render
Render with yafaray with 1 sun lamp
Render in blender internal GI.
Fg -10 sample
5 bounce
Factor 0.85
Attenuation distance - 4.000
no shading
Environment lighting - 4.0
This render I decrease the Attenuation distance - 0.600
Notice that the wall at the right is blue which is not the case but which is illuminated with the environment light.
It is very slow to render when I render it with 100 sample which still I see the noise
The bouncing light or GI, I think doesn’t goes far enough to illuminated the whole scene even if I increase the factor.
I also increase the sun intensity but the room is still dark
Its very slow to render glossy object in the scene so I remove all.

So here is my test. Hope this help to improve blender indirect lighting.:yes:![][IMG]](][IMG]

A lot of images don’t get displayed for me…
btw, are you using the blender render branch?

You didn’t mention the mesh emitter? around the ceiling which also seems to be giving out quite a bit more light in Yafaray than in the BI renders.

@Freeman. Yes Its the render branch rev 29611 32 bit.
@Seanser. The mesh emitter around the ceiling don’t play big role. Its mainly illuminate by the sun and the skylight. But blender environment lighting doesn’t illuminate well the scene.

I’m not sure how BI’s global illumination works but maybe an implementation of exit portals around the windows as you would normally use in other renders, Vray etc. could help in an interior environment like this.

Also, is the BI render tone mapped?

Sure it help if I put an area light around windows. I just want to make a comparison like it is. As for the ceiling emitter inside blender its already act like a meshlight. I just want to help improving it as they are working on it because I cannot get a good result with interior scene which lit with a minimum of light. And also about the render time and the noise.

if not already, please test with 1 single bounce and cache enabled

For what i know, the GI in blender (Render Branch) doesn’t support more than 1 bounce. Doing more than 1 bounce at this stage will produce bad results. I think i have read that the devs will tackle this post Durian, but i’m not 100% sure.

I try it at 1 bounce and cache enable but the noise is still there at 20 fg. However I don’t know why, just the ceiling is smooth and some part of the right wall. I disable all mirror material.

As someone said arealights in the windows would be a big help here as the GI adds a lot of light with the assistance of arealights.

The good news is that the arealights are now unable to burn out the scene if you have the smooth value high enough.

Over at Luxrender Xelptic created a “Red Room” for testing. I gave it a shot in Blender25 GI (rev.29557) - with some slight modifications, including area lights in the windows etc.

With IDL samples set to 64 — 50mins on Intel E8400, WinXP Pro 32.

I suppose that someone who knew what they’re doing could get more out of it

Here’s the blend file:


The main reason why Blender cannot even compete with any other render engine is that
it cannot efficiently render complex materials such as glossy surfaces or caustics.

The GI system between Yafaray and Blender Render Branch are very different.

Blender is more geared towards fast rendering results but not realism and material realism,
which in itself is not bad when you do not need the extra detail but look for visually pleasing
results for motion picture.

As it looks this is also their internal goal. If they want to explore this more towards
adding realistic material renderings in an acceptable render time is another question.
And if they have man power and time for that.

The new GI system would be able to at least goes a little ways to total realism if the bugs related to using full shading with multiple bounces is fixed.

I have said this again because right now the bug(s) that cause black pixels prevent you from really pushing the new GI system, and I do hope the fix isn’t by removing that shading option because of its importance for more realistic rendering.

Also, the per-tile displacement implementation seems to have never been finished since it wasn’t used for a lot of things in the Durian project.

If the Blender institute ends up saying these bugs will be fixed more quickly or incomplete features finished faster by way of donations then by all means make the donation button visible on the frontpage and let see how many here will donate, particulary since there will probably not be another movie project for six months once Durian is finished.

AD, no intention to be rude or offensive, but your complains about the current flaws of the render branch are really getting repetitive and boring to read. You’re totally right about the bugs, just be patient until Durian will be done, and devs will tackle them and then push the render capabilities further. Isn’t it already stated that multiple bounces will be fixed in the future? And of course, black pixels are a bug, so they’ll get fixed. Also a new shading system is on the horizon, and you know, shading as well as lighting plays a big role in getting photorealism. So…
Again, please don’t feel offensed.

Well, I just feel at times I need to be a broken record because I’m not sure of the percentage of threads here read by the devs. (like the ones related to the render25 branch because I know they don’t have time to read everything), and to raise user awareness of these bugs that have been in the render25 branch since the first month of its existence.

Don’t worry AD - I’m sure we’ll get there eventually.

You know, if you were to put a link to the donation page in your sig it would show up in 13824 different places on this forum.

And I’m sure throwing money at the problem couldn’t hurt.