I'm just curious - does it make sense to increase GI relative detail and GI sample quality over 2 with supersample prepass checked?
I've found that setting GI quality on 3 increases overall render times significantly without producing visible differences.
Or maybe it's because I'm not lowering GI blur radius together with GI detail?
This is indeed an interesting remark/question.
I don't know how well your understanding of TG's GI-system is and especially the GI blur radius setting, so if you don't mind I'll explain/show how to...shortly:
Start new scene, set camera near ground and add a sphere close to the camera and let it sit on the ground. Set lighting angle 90 degrees from left/right so you'll have a nice long shadow. Set renderquality to medium (0.6 - 0.7) and GI 2/2.
Render with blur-radius 0 and blur-radius 10.
What you'll see is that with radius 0 bright sharp-edged but smooth shapes will appear in the shadow.
With radius 10 the same detail is just 'blurred' (oh really? lol ;D) and therefore less visible.
That's the blur radius in a nutshell.
So when do you use what?
Unsurprisingly, this all depends on your type of scene.
In general virtually all images improve from GI 2/2 over GI 1/1. GI 3/3 takes indeed considerable more time to render. GI 3/3 or 4/4 or maybe even higher is something you can better use when features of the terrain are extremely displaced for example. It is then difficult for the render to sample all the displacements properly and this can result in black areas and other lighting artefacts. Increasing GI settings increases the "coverage" of the GI so that light and shadows will be calculated "correctly" in difficult areas.
The blur-radius then determins, well I think, how well GI details will be visible. Like I tried to explain above. So I think you can approach them seperately. However, like you said, if you choose to render with lower GI settings, when you're rendering grasses for example which render almost always fine with GI 2/2, you can reduce the blur radius to "exaggerate" the GI effect.
This is a bit of a philosophical approach of course and maybe here and there not really accurate technically.
Martin
So basically:
In order to have better quality image with increased GI samples and detail, you have to decrease GI blurring?
It makes sense as there are more GI samples, so blurring is not necessary.
At last in my theory ;)
No, I think that decreasing GI blur radius isn't the same as blurring images to get rid of grain or to mask/cover areas of low quality.
At least, that's what I think I understand of what you mean.
In my theory it just determines how sharp the GI details are visible. The GI detail settings just determine how accurate these GI effects are being rendered.
The blur-radius then does or doesn't blur these details.
Have you tried the scene-setup I propesed to check this out?
Martin
Not yet. I have a render in progress at the moment. I'll check that tomorrow and I'll do some experimenting with vegetation or heavily displaced terrain.
See these:
http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=9505;image (http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=9505;image) (GI 2/2)
http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=8736;image (http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=8736;image) (GI 4/4)
Though the lighting setup is somewhat different between these two (enviro light strength settings) the essential difference is in the GI quaility settings when it comes to filling dark traingle-shaped spots on the rock-formations like you can see in the first image. The second image has higher GI settings and therefore has less triangle-shaped black areas.
Martin
@TU - Your second example is too small. :D
@TU - Did you know...it's just a thumbnail?
Yes, second image is just a thumbnail, but first one is bigger and those triangles you mentioned are visible.
Btw. how did you generate this terrain?
Quote from: calico on August 09, 2008, 06:33:20 AM
@TU - Did you know...it's just a thumbnail?
Oh no I didn't....haha..lol...I first thought it was a kind of a cynical joke, therefore the ;D
I'll try to fix the link...
Quote from: PorcupineFloyd on August 09, 2008, 07:19:45 AM
Yes, second image is just a thumbnail, but first one is bigger and those triangles you mentioned are visible.
Btw. how did you generate this terrain?
It is done with functions, powerfractals and strata&outcrops shaders.
You can find the tutorial at ashundar, here:
http://www.ashundar.com/index.php?action=tpmod;dl=item342 (http://www.ashundar.com/index.php?action=tpmod;dl=item342)
I've never met/seen anyone who've read it ;D lol
Hope it helps, otherwise just ask.
Martin
So, let me be the very first person to read that thing :D
I mean - I'm honoured :P
Thanks TU! I love to see the differences so clearly, especially with your explanation. Thank you.
Hmm, the link still isn't fixed, TU... Could you fix it , pretty please?
Oops...I forgot already ;D sorry!
Here it is:
http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=8735;image (http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=8735;image)
Already had the time to test the blur-radius PorcupineFloyd?
Martin
Thank you, TU. Amazing difference.
I'm currently working on a render with a mushroom shaped terrain feature (build from an inverted crater shader and some power fractals) and those black areas, caused by the displacement, really are a problem. GI2 didnt do it so I'll try GI4. And if that still looks crap I'll try GI Joe.
Quote from: Tangled-Universe on August 11, 2008, 05:22:02 PM
Oops...I forgot already ;D sorry!
Here it is:
http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=8735;image (http://forums.planetside.co.uk/index.php?action=dlattach;topic=3217.0;attach=8735;image)
Already had the time to test the blur-radius PorcupineFloyd?
Martin
I did. However - not on a plain example, I've just made a render with 4/4 and blur radius lowered to 6 (and it looks good). However - I'll have to experiment with various GI densities and blur radius. Maybe I should do some plain renders and post them on the forum or even TG wiki?
Oh, by the way... this tutorial made by you is amazing! This node network is impressive :)
Thanks! Glad you like it and find it useful (hopefully) ;D
You don't have to try different GI densities. GI 2/2 is sufficient to determine the difference in blur radius settings.
I'd go for big differences (always) when determining the effects of settings.
Martin
I believe GI Blur is essentially good for covering deficiencies/inaccuracies in GI from low sample levels. You want to strike a good balance of sample levels and blur. In most cases 2/2 is probably good along with the default blur, but if you find the GI looks "blotchy" for example, you could increase the blur. The feature was however mostly introduced to allow animators to have a way to use GI while avoiding flickering between frames, so in still frames you may not need to change it at all.
- Oshyan