The above animation was rendered in Max9/VRay 1.5RC3. I am going to be switching the farm over to 1.5 final in the next couple of days and re-rendering this sequence. I know that it’s not advised to do this, but I can always convert back if I encounter problems.
I’m seeing a splotchy noise in my animation. I don’t think I’ve quite seen this noise effect before (it does not look like Irad noise), but I was wondering if it might be my secondary Light cache? Maybe something else? I would appreciate any comments/feedback/etc.
Settings:
Adapt SubD: Min -2; Max 7; Clr Thresh 0.2; Outline off; Rand Samp on
Irad: Min -3; Max 0; HSph 45; Interp 25; Clr 0.4; Nrm; 0.1; Dist 0.5; Delone
LightCache: Subdivs 400; Size 0.02; Screen; Passes 4
wow, there is many problems in your setup. I will only say that, having adaptive subdivision at -2/7 is not a good idea.
The flickering is caused by irradience map. Since you have a lot of moving objects you need to use irradience map for animations, which comes in a new version of vray. Otherwise, you cant use irradience map and lightcache on moving/deforming geometry.
Well, like I said, I’m moving this to 1.5 final right now so that I can use the new Irad animation mode. It was not available to me before SIGGRAPH, so other than QMC brute force I didn’t have a whole lot of other options. What else is wrong? Why is -2 7 bad? What should it be?
typicly undersampling an image has no advantages; usually 0/3 adaptive subdivision is enough. Decreasing color threshold will create more fine sampling.
woah.. thats alot of 3d people. Yeah standard IR map just wont cut it with animated objects. The new animation type should help you out alot. But you may need to revert to QMC as your 2nd bounce. As LC might become an issue.
So am I understanding this correctly: we can have both animated cameras AND objects (people etc) and use the new IRmap Animation setting to precalc an irradiance map solution WITH moving ojects?
I always thought that using any kind of optimised GI (Irmaps for instance) prohibited you from having animated objects because the Irmap would be changing over time and cause flickers.
gilpo: Olli96 is correct, the people are AXYZ, RocketBox, as well as a few Lowpolygon3D.
sea2stars: Yep, 1.5 final.
I have been pushing about 5-6 new animations through the new animated imap mode. It’s taken a while to tune the settings just right, I’m currently averaging about 3.5 hrs a frame. The brute force settings I was using with RC3 was nearly 10 hrs per frame, and these 3.5 hr frames look a whole lot smoother and flicker free. I should be able to post a sample in a day or two.
And just to clarify because I don’t think it’s clear if you weren’t at the meetings at siggraph, the new “IR map with animation”, is really just a new irradiance map for individual frames, which are then blended together across frames for the final image. That’s why it works with animated meshes, because every frame has it’s own GI calculation.
So here’s an update. It’s nearly final, but I need to increase the image sampler a tad more. It’s an adaptive 1,3 right now. I’m thinking that 1,5 should do it?
This gives me about 40 minute Irad press frame times, and about 3 hour final render times. I have noticed that the Interp Frames setting can add a lot of time to a render. Below are two sets of renders, the top set has a very low image sampler settings, the lower set has fairly high settings. You can see that as I change only the Interp Frames setting, the render times shoot way up (especially on the higher quality version).
At the Chaos party during SIGGRAPH, I remember Vlado saying that increasing the frame blending would increase render times, but I got the impression that the increase would be much less than what I’m seeing. I think he said something like 5-10%? I could be wrong tho. It’s still an amazing time savings over all. The above animation had about 3.75 hr frame times and there’s basically no noticeable GI flicker. In RC3, I was pushing 10 hr frame times and I still had some noticeable GI flicker.
Now we just need some good updates to the online manual. For example, I still have no idea what the different Path Samplers do. (Legacy, Latin Super Cube, Schlick) And, why do I have to specify the number of threads in the LC (Number of passes), why can’t it auto detect how many cores are available on the machine?
just wondering if you think that maybe changing the interp setting from 25 to 60 has helped smooth things out or it is mostly due to the new interpolation between frames ?
gonna have to get myself a dongle now to try this out
BTW adaptive 1/3 is very very high for an animation. I know you want it to be sharp but can you afford these rendering times?. we generally try to tweak the settings to get it under 30 mins a frame otherwise it takes too long. The quality with your settings is starting to look good though IMHO…maybe 30 mins just is being too optimistic?
It depends on the scene, I suppose; if you consider that with the interpolation frames set to 3, V-Ray is loading 7 irradiance maps at the same time, it has 7 times more data to process for each shade than with a single frame irradiance map.
Now we just need some good updates to the online manual. For example, I still have no idea what the different Path Samplers do. (Legacy, Latin Super Cube, Schlick)Mostly, leave the default here. In newer versions, only the Schlick sampler will remain, the others will be removed.
And, why do I have to specify the number of threads in the LC (Number of passes), why can’t it auto detect how many cores are available on the machine?Because, if that were the case and you were rendering an animation on the network on machines with different number of processors, each will produce a different result. This is the exact reason why the concept of light cache passes was introduced. In the beginning, there was no such thing, and light cache maps produced by machines with a different number of processors could be wildly different, even for the exact same scene… not good
It depends on the scene, I suppose; if you consider that with the interpolation frames set to 3, V-Ray is loading 7 irradiance maps at the same time, it has 7 times more data to process for each shade than with a single frame irradiance map.
I have also noticed that if you have larger irradiance maps (more irad points), it takes longer to process the blend. eg., If I render the irad set as -4 -2, I typically get 2.5MB map files. An Irad pass of -4 -1 creates maps about 9.5MB each. Now if I do a final render with each set of irad maps using same interp frames setting, the 9.5MB irad maps add a lot more time to render than the 3.5MB irad maps, about double in my test scene.
Mostly, leave the default here. In newer versions, only the Schlick sampler will remain, the others will be removed.
Ah, so the other samples will be removed in future versions? I had not heard that one yet.
As for using adaptive 1/3 for animations; I pretty much have to use 1 as a minimum, or my thin wires go missing in some buckets, and 3 is actually lower than I want for a high sample. Small parallel lines generate moire patterns even at 3 or 4.
Vlado: Is there any way you could add a Sampler Modifier to the VRay material? For example, say I have my Adaptive set to -1/2 and I make a material and set the Sampler Modifier set to 2, then when it renders anything with that material the sampler works at 1/4. I think that would be really cool!!! (this could also be a VRay Object property)
