Why IrMap over Brute force for animation rendering?

[iancamarillo]

What are the benefits of locking the noise pattern?

[davius3d]

thanks for the comments.
Yes. The scene only has an animated camera, so i would expect there to be huge complications if i started animating objects within the scene.

With regard to the flickering rosemary. If you look closely that both animations, the flickering is exactly the same on the rosemary plant. I think DOF is creating the illusion that it is flickering.
Also, the shadowing around the edges with the IM+LC seems to be more stable (if you look very closely). NOTE: you have to download the anims to your local machine to view.... Viewing it on Dropbox is rubbish.

That being said, I am certainly open to using BF+LC. I would MUCH prefer to use this method. I just need to figure out how to get render times down.
I will keep trying to find more resources on the technique.

I did download and can safely say that both render qualities are on par (based on my biased eyes). I didn't see at all any flickering as result of GI calculation on the Irr+LC method, Lele - could you point it to us? I don't mean this defiantly!! Not at all! But just like SSS, getting to "know" GI flicker (when very subtle) takes a specially trained eye. I see some flicker (on the oven bright spot) but this is present on both methods and I think it's due DOF.

Just as 4754simon said - I'd prefer MILES to use BF+LC everytime! But the time cost (and dollar cost related to it, specially if you use renderfarms) is not negligible - it can sum up to several thousand dollar in one job alone. The thing is - we WANT to use BF+LC, but it is just NOT as fast as IRR+LC where only the camera moves. In this regard Chaos can do several things like improve BF+LC even more to rival Irr+LC and/or improve IrrMap+LC even further to keep all options open. I feel that the Irradiance Map workflow was kind left in the corner in favor of BF+LC, but it still is a BIG win for Vray!

4754simon could please share also your settings for the IrrMap+LC and BF+LC scene? I'm curious how you set it up.

Cheers

[Rens]

I tend to do a lot of IM/LC for shots. BF does look better but with say 100-300 frames and just a moving cam the look difference is negligible and the time savings are huge. Especially if you have some way to automate setting up the precaching, like set the file and/or submit both render jobs.

I rarely ever move away from 100 subdivs, 5-10 interpolation samples, min max -3/-1 and LC samples set to 1500. If you don't need to change those settings the precaching process is quicker to set up. I agree that the IR/LC setup process would benefit from being simplified so it's a simple click of the button to switch, render, click, render again, or better yet one click to send off a render job with both precaching and render. That way you'd have the best both worlds easily accessible as IM/LC isn't going to be beaten time-wise for a while for those kind of shots.

[4754simon]

4754simon could please share also your settings for the IrrMap+LC and BF+LC scene? I'm curious how you set it up.

Cheers

@davius3D.
not sure if it sheds light on anything? I"m hopeless when it comes to understanding Vray settings. The only thing i changed for Brute Force settings ... lightcache was pushed to 2000. Usually its set to 1000 for my stills.

[AngelStudio]

I'am so confuse right now..
I never do animation for production work in vray only some tests at home .
First Light Cache ( Fly-through mode) Save it ,pre-calc Irradiance Map with Medium-Animation preset , finally render all the frames .
As i read now i only have to set LC+BF with "correct" subdivs and Retrace value and render all frames?
So i have a scene that takes 1.5 hours to render at 1280x720 , consider 600 frames i have to wait 900 hours ( 37 days) to render the animation?
Does the IrrMap+LC do the job in less time and same result ? Or can i make some adjusts in BF+LC method to get fast render times, like use denoiser and increase noise threshold in Image Sampler and DMC?
Hope someone can't explain i pretend to do some interior/exterior animations but with this render times it's crazy.

Thanks

[Jeffrey]

IrrMap + LC is much faster, period.

LC + BF is more or less a general 'it just works' setting. But it comes at a (render) cost.
Just throwing in more cpu power for the sake of getting it done is a bad advice imho.

There's definately much room for rendertime optimization with IrrMap + LC. Yes, it requires some fiddling with the Vray settings. But in the end this is good as it helps you understand how Vray works.

[tricky]

Right then, time to put BF/LC to the test with our latest production. We have always hindered some of our creative ideas by being a slave to pre-calculated GI solutions: no moving objects...no moving lights etc. This time we are being a bit more daring and even opening doors before passing through them!

Using BF/LC almost removes the technical breakdown of each shot where we pull a scene apart to render different things separately before comping back together later so that should save a good amount of time. I also like the fact that if a few frames aren't quite right, I can simply open the scene and render them again without trying to find the correct saved solution etc.

We'll see how it goes...

[4754simon]

I'am so confuse right now..

AHAHAHAH. Don't worry AngelStudio, you are not alone... HAHAHA. hence my ramblings earlier. I will leave your question regarding 'can i get faster render times using BF+LC'... because i have no idea.

Based on my tests the IM+LC takes half the time to render a BF+LC scene. However, keep in mind that i was ONLY animating the camera.
The only difference between the two techniques is that you need to create 'pre-passes' before you 'send to the farm'. This is certainly not a taxing exercise, however when you are tired, it is easy to make mistakes.

I recently did another short animation test... (please download to view)
https://www.dropbox.com/s/9b7q1ubnlt...eyway.mov?dl=0

This animation was rendered using IM+LC (1280x720... and rendered approx 7mins a frame on my small farm). It rendered quite cleanly except for a few flickers from bad geometry from the consultants model.
Also, you will notice towards the end of the animation, some glowing flies appear over the garage door material. This possibly wouldn't have happened if i had used Brute force, however if i had known that the 'flies' would have appeared, i probably could have spent more time on the material of the door. (one of those things you don't notice until you have finished rendering).

[Oleg_Budeanu]

Wanted to add my experience to this thread if you guys don't mind.
I am working now on a quite simple fly-through animation for an interior. The first test was done with camera path mode on, everything was quite good - the length of the animation was 1000 frames. Though at about 870 frame appeared a difference in sunlight! From one frame it decided to be brighter - affecting both lighting and specular elements. And it decided to be that way till the end.
So, i did not find another solution and ended up rendering every 10-th frame with multiframe incremental mode. And I ended up with last bucket stuck issue and had to remove some things from the scene. And render times went twice higher, though the quality of GI is twice better. So I think for short camera fly-throughs it is OK to use camera path. With something more complex - incremental mode. Did not even compare to brute force as time is crucial for me now...

[jackieteh]

"Question from william.morris Lele! You legend! Thank You!! I have been looking for a thread like this for so long!! I will definitely be giving these settings a go! I have avoid rendering moving objects for so long as the precalc used to take ages. One setting which always confuses me for animations with moving objects.....'lock noise pattern' should this be on or off?"

Hi, Lele, I would like to know what is the answer for this?

[grasshopper]

Is everyone using progressive image sampler for their tests? Between using different Gi methods and Sampler options we are getting a huge variety of unsatisfactory results for simple a animation with only a moving camera.

BF+LC just seems too slow and progressive aa is introducing to much flickering in texture surfaces and small detail. What would the recommended noise thresholds be for the progressive image sampler and Global DMC when producing an animation or should we be using buckets for better stability?

Picking through other peoples findings and recommendations for render settings is ridiculous. Where are we with getting some proper updated documentation?

[grasshopper]

I am also finding that pre-calcing the GI with IM+LC (save to file and rendering a single frame) can take far longer, sometimes 5 times longer than not pre-calcing it. Whats going on here?

[^Lele^]

"Question from william.morris Lele! You legend! Thank You!! I have been looking for a thread like this for so long!! I will definitely be giving these settings a go! I have avoid rendering moving objects for so long as the precalc used to take ages. One setting which always confuses me for animations with moving objects.....'lock noise pattern' should this be on or off?"

Hi, Lele, I would like to know what is the answer for this?

Sorry, i entirely missed the replies here.
"lock noise pattern" is best OFF for animations.
Assuming some visible noise is present in the final render, if the lock was active, the noise "placement" would stay constant on the image plane, and the animation would look as if it swam under the noise pattern.
With the lock off, noise placement will change per frame, and noise will be the usual sizzling affair we know (and loathe. XD).
The unlocked behaviour is apparently more pleasant and readable to the human visual system, so in general the default for the locking of the pattern is set to OFF.

[sharktacos]

So I get why the light cache subdivs would need to increase with image size since the little "blobs" the light cache uses would need to be distributed over the large screen space. I'm having a hard time understanding why the light cache retrace threshhold would increase exponentially with increased image resolution. As I understand it the retrace threshhold refers to the distance of an object to the camera. So why would that distance be effected by a larger image resolution?

[^Lele^]

So I get why the light cache subdivs would need to increase with image size

Not necessarily, why do you say that?

since the little "blobs" the light cache uses would need to be distributed over the large screen space.

Yes, but it's not done daftly, and most of the potential issues are taken care of by Leak Prevention and retrace

I'm having a hard time understanding why the light cache retrace threshhold would increase exponentially with increased image resolution. As I understand it the retrace threshhold refers to the distance of an object to the camera. So why would that distance be effected by a larger image resolution?

Likely because now you're shooting a lot more, smaller LC samples, and the leak prevention goes into overdrive.

It'd be good if you could show what issue you're having which forces you to raise LC subdivs for a still when resolution changes, as it should really be resolution independent by design.