Animation with GI that doesn't flicker

[vlado]

Indeed in theory the photon map may be somewhat more stable, if the lights do not move. You can use it only for secondary bounces though - for primary bounces it would require way too high settings. Since V-Ray also supports photon maps, you can probably try that too. This would be only practical for interiors too - photon maps do not work well for exteriors.

Best regards,
Vlado

[duke2]

Ok so they're no good for animated scenes.

[vlado]

Here is my attempt with the scene at the start of the post, done with a beta version of SP3. It's rendered with irradiance map (animation mode) with medium preset, 50 hemispheric subdivs and 20 interpolation samples, no detail enhancement; light cache with 200 subdivs, no prefiltering and 10 interpolation samples.

http://www.spot3d.com/vray/images/st.../test_walk.avi

Best regards,
Vlado

[duke2]

Neat. The noise looks more stable.

[Morbid Angel]

what I would say about this test is, it needs to be done in a closed space, where gi bounce is more obvious. In such conditions, the flickering might be much less noticeble. And also a higher compression on the video will help.

[duke2]

what I would say about this test is, it needs to be done in a closed space, where gi bounce is more obvious. In such conditions, the flickering might be much less noticeble. And also a higher compression on the video will help.

Less noticeable? The idea of the test was to show it in conditions where it's as noticeable as possible More often than not people get flickering in certain parts of their animation rather than the whole thing, and I was trying to stress VRay as much as possible by feeding it a more complex problem than your average "door opens" or "car drives by".

Any suggestions as to what a redone test scenario should contain are welcome though - i'm just unclear on what you're saying I think.

[Morbid Angel]

basically, when you have an open space such as in this test, the bounced gi is "less" in a sence of blending between direct and indirect light. Where as, the skylight being as a direct gi ray will illuminate most of the scene, and therefore will contribute a lot to an indirect blending. It may sound confusing, but this test is actually not showing gi for animations too well.

What I was suggesting is to take a scene where a direct ray weather gi or other is minimized and the rest of the solution is left purely upto indirect gi, then this would show very well how the animation is being calculated for gi blending.

[duke2]

basically, when you have an open space such as in this test, the bounced gi is "less" in a sence of blending between direct and indirect light. Where as, the skylight being as a direct gi ray will illuminate most of the scene, and therefore will contribute a lot to an indirect blending. It may sound confusing, but this test is actually not showing gi for animations too well.

What I was suggesting is to take a scene where a direct ray weather gi or other is minimized and the rest of the solution is left purely upto indirect gi, then this would show very well how the animation is being calculated for gi blending.

Ah ok, so an inside the box with a pinhole in it type deal. I might revise the scene tonight to show an internal thing with a wood-brown floor, a couple of windows and a walking dude or two.

[duke2]

Ok i've come up with a new test scene. I tried the settings Vlado used and it's indeed flickering but it will be good to see it in SP3



http://www.dukecg.net/anim_test2.mov

http://www.dukecg.net/anim_test2.zip

[vlado]

Ok then, will try it when I get to work a little later today.

Best regards,
Vlado

[vlado]

Here is what I got (the frames in the .mov are compressed with lossless PNG compression):

http://www.spot3d.com/vray/images/st..._imap/test.mov

The settings in the scene were too high though, so I've reduced them a bit:

http://www.spot3d.com/vray/images/st...nim_test2a.rar

I'm rendering now again with some camera move also.

Best regards,
Vlado

[duke2]

Here is what I got (the frames in the .mov are compressed with lossless PNG compression):

http://www.spot3d.com/vray/images/st..._imap/test.mov

The settings in the scene were too high though, so I've reduced them a bit:

http://www.spot3d.com/vray/images/st...nim_test2a.rar

I'm rendering now again with some camera move also.

Best regards,
Vlado

Wow. That's flawless. Would you attribute most of it to the time independent noise now working with animation mode? And when's SP3 out dammit!

[vlado]

Would you attribute most of it to the time independent noise now working with animation mode?

Yep, entirely. There are no other changes to that code.

And when's SP3 out dammit!

As soon as we can get it... there are still a few things to finish for the mesh light and incremental irradiance maps in DR mode. You can (almost) always get a newer build if you email me though.

Best regards,
Vlado

[Morbid Angel]

duke2 I had a problem opening your scene, the settings would not transfer to newer vray. But as vlado made this test pretty conclusive. Vlado what were the render times like?

[vlado]

Here is another one with a slight camera move also; for this one I had to turn off the "Use light cache for glossy rays" option as it was causing flickering. Again frames are compressed with lossless PNG compression.

http://www.spot3d.com/vray/images/st...imap/test2.mov

Best regards,
Vlado