Static prepass render settings

[sharktacos]

First question:
Rendering a flythrough animation of a static scene (i.e. only camera animation) of an outdoor environment, I ran into some conflicting information:

1) For the Light Cache, the manual says:
"Fly-through - this will compute a light cache for an entire fly-through animation, assuming that the camera position/orientation is the only thing that changes. The movement of the camera in the active time segment only is taken in consideration. Note that it may be better to have World Scale checked for fly-through animations. The light cache is computed only at the first rendered frame and is reused without changes for subsequent frames." (source:
http://help.chaosgroup.com/vray/help...lightcache.htm)

2) However I read this in the "Rendering a Walk-through Animation" tutorial:
"Keep in mind that when you use the World scale, you must make sure that your scene is not too large (e.g. with a large ground plane) or that the light cache samples are not too small, otherwise you can run out of memory for the light cache. This is not such a problem for the Screen mode, since surfaces that are far away from the camera will get fewer samples anyways. For exterior scenes, it is recommended to use the Screen mode always." (source: http://help.chaosgroup.com/vray/help...map2.htm#part1)

So for a calculating the LC in flythrough mode for an outdoor scene should world scale be checked or not?

Second question:
The scene has grass in it (Vray fur) which is prone to flicker. Would it thus be advisable to have prefilter on?

Our workflow is:
1) render LC in flythrough as a single frame (LC both in primary and secondary GI), save it to disc
2) read the LC from file and render primary IR and secondary LC, saving the IR to disc in multiframe incremental mode.
3) Set IR to "from file" and LC off and render the sequene

The manual says "The light cache file does not include the prefiltering of the light cache; prefiltering is performed after the light cache is loaded, so that you can adjust it without the need to recompute the light cache."

So does that mean it would be best to have prefiltering OFF for step #1 above, and then turn it ON for step #2 ?

Thanks for the clarification!

[sharktacos]

Another question, regarding the IR precalcuation:

If we are precalculating the IR by rendering the first and last frames using "multi-frame incremental" with "use camera path" on, and then saving that IR to disc. Would it work sense to have "enhance details" on here? As I understand it, enhance details uses a brute force approach for the details and an IR approach for the other parts. So will enhance details work when rendering in steps like we are? or does one need to render every frame for it to be accurate (in which case we could not use it here)?

[Svetlozar Draganov]

So for a calculating the LC in flythrough mode for an outdoor scene should world scale be checked or not?

For fly-through animations it is better to use World-scale option which will ensure that the samples are equally distributed over the objects in the foreground/midground/background - this is very useful for example if at the beginning of the animation important objects are in the background but at the end they moved in the foreground.

However if in the objects in the background never gets to midground or foreground it is better to use Screen-mode because Vray will shoot too much samples on those objects which will occupy a lot of RAM memory and at the end those samples won't be needed at all.

Like always everything depends on the scene - but in general World-option in light cache is preferable if camera moves and reaches to objects which were far away at the beginning of the anim, otherwise it is better to avoid it.

The scene has grass in it (Vray fur) which is prone to flicker. Would it thus be advisable to have prefilter on?

Prefilter will help to smooth out the Light Cache solution and it will produce cleaner animation - please note that every smoothing/blurring filter leads to loosing some finer-details.

So does that mean it would be best to have prefiltering OFF for step #1 above, and then turn it ON for step #2 ?

It doesn't matter if Prefilter is ON or OFF for the first step - I would say it is better to keep it always ON if you want to use filtering cause otherwise you have to keep it in mind that it must be turned ON for the second step.

So will enhance details work when rendering in steps like we are?

Yes - it will work. Like you said Enhance Details use Brutte Force approach which is calculated during the actual rendering, so it doesn't matter that you have a Irradiance Map which is calculated at each 10 or 100 frames , at later stage when you are performing the rendering Vray will pass trough every frame and then it will calculate Brute Force solution for those areas.
As a side note I would like to point out that if you had Irradiance Map which is calculated with details enhancement ON and at later stage you decided to turn it off the final result won't be accurate because Irradiance Map won't contain any information about the detail-enhancement areas.

[sharktacos]

Thanks very much for your detailed response, it was very helpful and clear. In our particular case the camera moves are pans rather than zooms so things in the background stay in the background rather than getting closer. Given what you have said I would say it would be best for us to have screen space on (i.e. world space off) for the LC.

An alternative approach that I am considering is instead of
Our current approach: rendering with IR/LC with detail enhancement and filtering on the whole time (precalulated as described in my first post)
would be instead this
Alternate approach: render with IR/BF with detail enhancement off.

I have heard that brute force in the secondary GI engine is good for outdoor scenes rather than LC. I'm thinking too that it could be redundant to use BF in the secondary and use it by having detail enhancement for the IR. So we could thus turn off detail enhancement and consequently get full coverage in our saved IR map for primary GI as you explained.

Which of the above two approaches (IR/LC vs. IR/GI) would you say would be the better one for an outdoor scene? Are there advantages and disadvantages to both that I should keep in mind?

thanks!

[sharktacos]

Also there is the option for LC "Use light cache for glossy rays." If the LC is being rendered in fly-through mode and saved out as a single frame (as described in step #1 of my first post above). Would it work to have this on? That is, when we later render the sequence using the saved IR map (that has the LC incorporated into it) and thus render with LC off in the secondary (since the LC map has already been based into the saved IR map) would having "Use light cache for glossy rays" on when we had originally calculated the LC have any effect at all?

I'm thinking it would have no effect at all... is that right? Sorry, thinking about this is making smoke come out of my ears

[Svetlozar Draganov]

Both approaches IM+LC or IM+BF have proven to be a very good solution for exterior renderings. The both produce a good quality image in a very short time with the following differences.

In terms of speed: IM+LC is usually is faster than IM+BF, especially when LC is used for glossy rays calculation. There are a few cases when IM+BF could be faster than IM+LC but they are very rare situations, and they are usually related to a scene which could produce infinite amount of GI-bounces. These for examples are scenes with a lot of trees - where a single LC-ray could travel forever - in such scenarios LC is usually a lot slower than usual.

In terms of quality - IM+BF will generate more realistic result with better details, better contact shadows and more accurate solution than IM+LC. However the result will be noisier because BF doesn't perform any interpolations over samples like LC does and if you want to achieve the same noise-levels you have to increase BF-bounces a lot which will leads to even longer render-times. BF also will produce darker-results in hard-to-reach regions - these are regions for example where a lot of bounces are needed in order to produce correct result.

Light Cache for glossy rays is not saved neither Irradiance Map nor in LC map - this option could only be used when LC is calculated on the fly but it has impact even if IM is loaded from file. So if you want to use it LC for secondary bounce should be always ON even if you already have it saved in IM. This means LC is calculate twice - once for IM and another once again for glossy rays but since in both cases LC speeds up the calculations a lot the final rendering of scenes with a lot of glossy effects will be much faster than if you have it calculated only for IM.

[sharktacos]

Brilliant. Thanks Svetlozar, this is very helpful!
Here's a related question:

Dome Lights vs. Environment GI color

For an outdoor scene where we simply want to have a blue colored GI bounce from the sky, would it be better to use a dome light, or the "GI texture" slot of the environment rollout of the render settings instead? In both cases this would simply be a simple color, and not an HDR texture map.

My understanding is that the more a scene is illuminated, the less the GI needs to work, and so by that logic it would seem a good idea to have a dome light illuminating the scene. However dome lights tend to have a lot of flicker and require very high sampling rates which slows down the render time. So despite that theory (more direct illumination will speed up GI) it seems that having the environment color for the GI produces faster and cleaner results. In the tests we ran, it seemed that adding a dome light (even when we had "store with irrandiance map" checked) actually slowed down the render times rather than speeding them up.

That might lead one to ask: Then what good is the dome light then? My understanding is that the dome light would be advantageous when doing HDR because it works more efficiently than having the HDR in the GI environment slot would. So I would tend to only use it for HDR and otherwise prefer the environment color for creating simple sky illumination.

Is the above accurate? Could you comment on the advantages/disadvantages of using dome lights vs. the "GI texture" slot of the environment overrides please? Are there other general strategies one should keep in mind?

[Svetlozar Draganov]

The main mission of Dome Light is to provide a smooth illumination without using GI , and to provide Image Based Illumination.
Dome Light is very effective and produce a very good results with HDRI image and this is why it is recommended to always use it if you intend to illuminate the scene with HDRI image.
The reason for the noise and the flickering is because Dome Light is constructed by many small VrayLights and the actual number of those lights depends on DomeLight subdivisions - with lower subdivisions means less lights which are calculated faster but the final render is usually noisier , higher subdivs means using more lights cleaner image and longer render times.
More direct illumination less GI work theory is true in general but if you have a scene with a million lights this is not entirely true because Vray have to calculate each of these lights and this will take a lot of time and it would be better to delete most of the lights and leave GI to do the work.

For a scenes with simple color it would be better to use Environment map than Dome Light , but if you would like to use image-based-lighting with HDRI it will suggest using VrayDomeLight.