Ok, I’ve seen so many topics about this in the last 6 months, I would like Valdo to set it straight. What is/was the reason for developing the Animation IRR setting?
I’m experimenting with this again today… I still don’t like it, I think. Now with DR working on animation - wouldn’t just running a multiframe IRR be faster? Even with Nth frame on 1 for moving parts? Is the result with the IRR interpolation THAT much better? (Seriously, Valdo, I’m asking). What is the benefit of animation pass, really? The IRR maps calculated on different machines don’t match, correct? And then what about the 2nd bounce in LC or Brute Force? This is different on each render node? So between the blocks of rendered frames it is really blending the IRRs? So technically instead of flickers we should be getting “waves”? And then if you turn up the interpolation frames - gets even slower…
But if you do IRR mutliframe with DR, you wouldn’t have any of this blending “waves”? Better chances for flickering with low IRR setting - but faster and technically cleaner in the end?
From what i’ve found - if you only have a small part of your scene animated/moving, it’s not worth it and works out slower. But if you have a massive amount of constant motion (facial/general in depth stuff thats a bit more complex) it works out a little better.
At the same quality settings, instead of sharp flickers it does produce waves at the expense of speed (on most architectural/large open stuff, the waves are still noticable anyway, so you may aswell just ramp the quality right up as you’ve found)
On areas with lots of motion, the animation distracts from the waves where the flickers would still be noticable.
The best way to do if you have only a small area of animation is pre-calc an irradiance map for the background, hide all moving objects, then use a v-ray sphere fade with the animation mode over your moving objects and comp them together.
It isnt much better than brute force, but it helps some.
Well this thing we are doing now is all moving parts, so I thought I’d give it a try. But it seems to have stuck/crashed in the middle. Was running fine, then all of a sudden its a patch were none of the machines will interpolate the frame and start rendering. Weird.
Thanks for the comments so far. So currently we are 3-0 against animation mode? Anyone else?
I’m no expert but it seems to work much better in removing flickering in exterior scenes than it does with interior indirect light scenes. I also think that using this method produces slower render times than saving out nth frames.
I’ve used the high animation preset for the irrmap (animation mode) + LC set to 1000 subdivs (single frame mode).
As you can see it has a smooth flickering caused, i think, by the changes in the light intensity. The windows size increases in time allowing more light enter in the room. The sun intensity multiplier is constant.
It looks like the lightcache setting is causing your problem. I tested a few scenes and for animations the setting has to be quite high to prevent the crawling effect. 2500 to 3000 should fix it but I switched to brute force for secondary. It was faster with better results.
1000 subdivs for the light cache might be too low to produce a stable light cache result, like TRGraphics said, 2500 or 3000 will work better (or brute force for secondary bounces).
Is it possible to merge IR prepasses? I have a scene with animated objects that I want to render from multiple viewpoints… if I could reuse the IR-prepass data for all the camera angles I’m sure it would speed up a chunk of the prepassing.
Also, can I just carlify that for the prepass render, I should leave secondary bounces brute-force/LC enabled… and in the final render pass, I switch off secondary bounces?
Does anyone know the answer to this? I’m about to start a large render job, and if there’s any way to reuse the cached IR prepasses between cameras it would be good to know about it before I start !
I can’t think of an immediate reason why this should not be possible using the irradiance map viewer, but it would be best to make a quick test (on a bouncing teapot or something) to confirm.
Wow, if that’s the case, that’s good news. I’ll give it a go and report back.
It makes me think it might be a good development to add an “incremental add to prepass” option in the IR calculation for scenes that have a fixed animation, but multiple cameras. After the first camera prepass has been calculated, subsequent camera prepasses would preload any existing IR-map (for the current frame) and then only have to calculate the missing samples. The only downside I can think of is excessively large prepass files…
I’m not sure how useful this would be to the wider vray community, but for my current job this would be a great timesaver.
For short film projects this would come in really handy - from what I gather a lot of people that use GI based rendering just have each scene with all the animation in and then a number of different cameras in it to edit the shots together from.
I know we’ve done a few interior animations for a buildup that had 2 or 3 cameras that needed rendering too.
