GI baking and mixing render modes

[Gecko]

hello,
I was wondering it it is possible to reuse some already baked GI infos in another pass : for example using the BG baked GI infos to speed up the pass chars which is calculated in brute force or Irradiance map single frame mode.
If yes, is there a specific work-around to achieve it ?

Regards

gecko

[Zdravko]

Please note that the GI rays do not create samples for the background.
What are you trying to achieve?

[Gecko]

Hello,
i am looking for a workaround that would help me to achieve this :
BG = ground + trees + buildings (all non-animated assets)
Chars = Characters + animated props

Bake GI on BG => Baked BG GI => Rendered BG
Render Chars using the baked BG GI for the BG Gi influence on the Chars (either Irradiance map or Brute force settings) => Rendered Chars
Then composite rendered BG + Chars as usual.

Gecko

[vlado]

It might work with the light cache as a secondary engine, but I haven't tried to see if there won't be any issues. In that case it might be useful to turn off the "Use irradiance map" option in the materials on the moving objects.

Best regards,
Vlado

[super gnu]

http://www.chaosgroup.com/forums/vbu...689#post553689

[Zdravko]

Here is the approach again:
- hide all animated geometry in the scene
- set the GI to Irradiance map and Brute force
- enable "Use camera path" option. Make sure you are using high settings for the Irradiance map as you are going to calculate it only once for the "BG" part of the scene
- enable "Don't render final image" option
- render the animation and save the Irradiance map. (you can also use auto save option)
at this point you have the GI baked for your BG

- disble "Don't render final image" option
- set the Irr map form file and load the file. Leave the secondary GI engine to none.
- unhide the hidden objects and in their materials disable "Use irradiance map" option
- now V-Ray will use Brute force for the the moving objects, the BF subdivisions however will be controlled by the subdivisions of the Irradiance map
- You can also use secondary bounce for those animated objects. Here you have a few options:
to use Brute force which will be slow but accurate, use Light cache which will be fast and accurate but you might get flickering or fake the secondary bounce engine by using V-Ray Ambient light + VRayDirt.

I hope this helps.

[Gecko]

Hello,
Thanks a lot for the help , i will try this workaround !!
regards,
Gecko

[super gnu]

this workflow has a problem as far as i can see. you are suggesting to calculate the gi for the static objects with primary and secondary bounces, but then the moving objects with only primary bounces. this means the static elements will have multibounce GI and the animated objects will only get a single bounce. this means the moving objects will come out much darker and contrast than the background.


to get this method to work properly you must use imap and lightcache for the background, save them both, then load them both for rendering.

the background will only use the imap, which is already based on the ligthcache's secondary bounces.

the moving objects will use brute force for their primary bounces, but those gi rays that hit the static geometry and take their secondary lighting values from the static lightcache that is "draped" over the scene.

since the lightcache is precalculated it will not introduce and blotchyness to the renders.

you will not have any secondary bounces from rays that originate on the animated geometry and hit other parts of the animated geometry, since it was hidden when the lightcache was calculated, however in most cases this is a negligible contribution.

afaik, this is the only way to get a decent result.

edit: i guess you could use brute force for secondaries in both the prepass and the render, but if youve ever tried to do a bf/bf render youll know this is a recipe for extreme rendertimes, even if its only apparent on the moving elements.

[refract]

"edit: i guess you could use brute force for secondaries in both the prepass and the render, but if youve ever tried to do a bf/bf render youll know this is a recipe for extreme rendertimes, even if its only apparent on the moving elements. "

Just chiming in here... I was under the impression that if you do brute force, you don't need to do a prepass, or have any secondary. ? or no?

[Zdravko]

this workflow has a problem as far as i can see. you are suggesting to calculate the gi for the static objects with primary and secondary bounces but then the moving objects with only primary bounces.

I mentioned that you need to use Secondary bounces when calculation the animation with the moving object. "- You can also use secondary bounce for those animated objects. Here you have a few options:
to use Brute force which will be slow but accurate, use Light cache which will be fast and accurate but you might get flickering or fake the secondary bounce engine by using V-Ray Ambient light + VRayDirt."

to get this method to work properly you must use imap and lightcache for the background

For the BG you can use either IM +LC or IM+BF depending on whether you are rendering exterior or interior. In both cases the the Irradiance map will store the secondary bounces.

I was under the impression that if you do brute force, you don't need to do a prepass, or have any secondary. ? or no?

Although the BF can not be reused when set as primary bounce engine. It can be stored in the IM if used as a secondary bounce engine, since the IM uses it for its own calculation.

[Gecko]

Hello,
Been away from this project for a while but i am testing this workflow right now, and one of my major concern is that the FastSSS2 seems to be using the IM by default and i haven't seen any way to counter that (and on this project i am forced to use sss on the animated pass).

So far, that method seems to work, i still have to check how to tweak efficently the IM parameters to tune the brute force on the Characters Pass, check the glossy reflections behaviours and so on.

i will try to bring more details soon.

Regards

[Gecko]

After some testing it comes out that while the final image looks nice and is pretty much noiseless , the render element Raw GI that i am willing to use in my compositing workflow comes out with a lot of noise: is there a way to filter this so i could get the same result as the final render (i have to keep my render times pretty low and tweaking the DMC sampler brings high render times (and doesn't bring more quality in the final render)?

i have come around another thing:everything looks correct (besides the noisy BF GI on the chars in the render element) , though if i set the specular/reflection ON ( raytracing reflections or not), it seems that the shader is then trying to use the Imap: it brings another artifact on the RawGI render element ( same artifact/problem with SSS).
Is it possible to implement some "use irradiance map" switch in the sss shader , and force the shaders with specular/reflection to use Brute force ?

To summarize, the workflow built this way doesn ' t seem bad and render times are somewhat what i expected , the noise on the characters is acceptable while using relatevily low presets, though it seems impossible to rebuild the final image with the render elements without bringing a lot of noise and some artifacts.


Regards

Gecko

[Zdravko]

Unfortunately you can not utilize this workflow when working with SSS, Blend and 2sided materials for the moment. But a material with specular/reflection shouldn't be using the IM. Can you attach an example scene, it seems to work fine here.
About the noise in the GI pass, you will need to increase the BF subdivs for the primary bounces which are controlled from the subdivisions of the IM and for the secondary from the BF settings.