Shadow gamma shift in animation

[digitalis99]

I made my first long-ish animated sequence using solely TG2.  I'm pretty pleased with it overall, but I have one major problem.  The rock wall in the shade shifts gamma pretty severely during the middle of the sequence.  I read up on the problems with GI and animation prior to sending this job to my farm, so I used a GI blur of 800 as someone here recommended.  Either that's not enough, or I can't use GI for this sequence.  Any ideas how I can get rid of that?

http://vimeo.com/33327856

[digitalis99]

Another problem I almost forgot.  That white band directly above the waterline is set to have an altitude constraint with a variation of .1m.  It varies closer to 7-9m.  Yes, I do have "Use Y for ..." checked.  Any idea why it floats up and down the canyon wall?

[Dune]

Did you use the settings in the surface shader or blended it by distribution shader? The last is better, set to 'final'. But of course, if you use a plane for a lake, it would shift over distance from 0/0/0. If you use another planet or sphere for water, you could also use a distance shader with a camera in the center of earth, and black and white set at water's level from the center + fuzzy area.

[MKE]

I still have the same problems with shadows in my new animation but finally I was able to improve the shadow quality from "horrible" to "quite bad". Some things I found out by trial and error:
- nearly all knobs you can turn in TG2 change the shadows but I didn't find the ONE knob that improves it.
- an extreme high GI blur radius doesn't really improve the shadows and sometimes creates very funny things. I had one completely black shadow while all the other shadows kept their normal behaviour. I reduced GI blur radius from 200 to 16 now.
- shadow quality improves with GI sample quality but there is not a stable good shadow quality if you choose GI sample quality just high enough. I ended with a value of 10.
- shadow flickering becomes worse on bright surfaces, e.g. snow.
- I installed a second sun at 90° with 1/10 of the strength of the normal sun. This brightens up the shadows a bit and reduces flickering.
- Like you I also used an DEM as height input and it seems that TG2 is very sensitive to the very smooth surfaces which result from the pure DEM without fractal detail. Now I use some additional roughness and my shadows become more stable. My values for fractal detail are: fractal amount: 0.05, fractal scale adjust: 2, fractal variation: 0.1, fractal roughness: 3, fractal flow factor: 0

But still, flickering shadows are THE problem in my new animation.

Kind regards,
Martin

[cyphyr]

I simply don't use GI in animations any more (well mostly ) at all!

Using a fill light setup as simple or as complex as you need/like will sort out all the GI flicker issues.

Just remember to turn off all the lights tick boxes. The more lights you use the nearer to a GI "look" you'll achieve but at a render cost (roughly a four light fill setup will render as quick as the default GI setup)

Hope this helps

Cheers

Richard

[Oshyan]

We'll have some improvements in an upcoming release that should help address the GI issues in animation. In the meantime a moderate GI detail, high GI sample quality, and moderate-to-high GI Blur (e.g. 50) is your best bet if you want to continue using GI. In the case of your animation I suspect it may have been the reflective surface very nearby that was creating quite variable results for each frame in the GI solution.

- Oshyan

[dandelO]

Oshyan, is the 'blur radius' based on pixels? Metres? Arbitrary number relative to another parameter?

I often wonder this and when I see this type of post, how do values of, say, 50, 500 affect the output, I'd imagine GI blur radius would be based on pixel radius but will this affect different sized renders? If I made a test render at low resolution and there was enough blur to smooth things out over the timeline sufficiently, does this number translate with increasing the render size or, should you increase blur radius relative to the resolution change?
I think it might be a pointless question for now since the GI is really erratic per frame but it'd be nice to know.

[digitalis99]

We'll have some improvements in an upcoming release that should help address the GI issues in animation. In the meantime a moderate GI detail, high GI sample quality, and moderate-to-high GI Blur (e.g. 50) is your best bet if you want to continue using GI. In the case of your animation I suspect it may have been the reflective surface very nearby that was creating quite variable results for each frame in the GI solution.

- Oshyan

Yeah...when is 2.4 hitting the shelves?

[digitalis99]

Did you use the settings in the surface shader or blended it by distribution shader? The last is better, set to 'final'. But of course, if you use a plane for a lake, it would shift over distance from 0/0/0. If you use another planet or sphere for water, you could also use a distance shader with a camera in the center of earth, and black and white set at water's level from the center + fuzzy area.

WOW...

I was using the settings in the surface shader.  I changed to using a distribution shader and got the results I was looking for.  Why the discrepancy?  Is it because the final Y value of a particular point isn't yet calculated in the surface shader, but the distribution shader can wait to get the final Y until after the total displacement is calculated?

Whodathunk.

[Tangled-Universe]

If you use the advanced search and search for posts by Matt or me (me basically repeating him) you'll find some in-depth answers to why this discrepancy is.

[MKE]

... I suspect it may have been the reflective surface very nearby that was creating quite variable results for each frame in the GI solution.

A reflective smooth surface is bad for flickering shadows, indeed. But it's strange that shadows nearby the camera are quite stable (although the viewing angles change more from frame to frame). Really bad are shadows from far away, in the background of the image. When approaching them, the flickering becomes less and when you're mostly there, it looks quite good.

I can't wait to test the next update... 

Greetings,
Martin

[Oshyan]

Oshyan, is the 'blur radius' based on pixels? Metres? Arbitrary number relative to another parameter?

I often wonder this and when I see this type of post, how do values of, say, 50, 500 affect the output, I'd imagine GI blur radius would be based on pixel radius but will this affect different sized renders? If I made a test render at low resolution and there was enough blur to smooth things out over the timeline sufficiently, does this number translate with increasing the render size or, should you increase blur radius relative to the resolution change?
I think it might be a pointless question for now since the GI is really erratic per frame but it'd be nice to know.

See Matt's previous post here:
http://forums.planetside.co.uk/index.php?topic=6350.msg68308#msg68308

- Oshyan

[dandelO]

Thanks, Oshyan.