Light cache, glossies and fly-through

[Frances]

Hi,

I've been using 1.4703 and learning about/testing light cache. It took me awhile to figure out that LC is stored in the IR map. I came across that here on the forum. I've noticed that when glossies are stored in LC, that a scene with lots of glossies will render faster with the LC in use rather than having secondary set to None.

I realize that Fly-through is meant for a single animated camera, but it apparently works for multiple cameras too? With LC enabled for Secondary Bounces and read from a file created in Fly-through mode, the test image rendered in 3min 14sec. With Secondary Bounces set to None, the test image rendered in 5min 48sec. Of course the quality of the glossies depends on the LC settings. With filtering and using LC, the glossies were nearly as nice at a render time of 3min 38sec.

Anyway, you can tell the difference on the side of the fridge at the right. Oh, and the glossies are all sparkly because I rendered with Adaptive QMC AA. They are much smoother with theAdaptive subdivision method.

[flipside]

Hi fran, not sure what youre question is, but here are some thoughts about the LC and glossies

IR map calculation uses the info from the light cache, so afterwards it is stored in the saved IR map. It's the same if you use QMC for secondary bounces. Once you save the IR map and want to rerender with it, you can turn off secondary bounces in the GI settings.

Because the lightcache has an option to use lightcache for glossy rays, there will be problems if you turn off the LC GI engine. The glossy material info is not stored with the saved IR map like the second bounces are. So if you want to use a saved IR map AND the "use lightcache for glossy rays" option, you need to let the secondary GI engine set to lightcache.

Altough that option can speed up rendering, I usually don't like it as the quality of the glossies is reduced in many cases.

For your sparkling glossies, this can be easily fixed. When using adaptive QMC AA, you need higher subdivs for glossy effects than when you're using adaptive subdivision AA.

For adaptive subdivision, glossy subdivs of 3 to 10 are really good enough. With adaptive subdivision AA, you need 20 to 40 subdivs for good quality. But don't worry, rendertimes will be even faster with the adaptive QMC AA if you have lots of glossies in the scene (like in your scene).

The QMC sampler should also be used to control the image quality. Noise threshold should be 0.005 or lower (0.001). Beware that changing settings here will also affect other QMC related parameters (DOF, MB, area shadows, irradiance map etc...)

kind regards,

wouter

[Frances]

Thanks wouter. I will try your suggestions. I have been doing a lot of research, here and with the online help (which I can no longer access since the site change).

My situation with this particular scene is that I'll render about 10 views at medium quality so the client can choose 3 (or more) views for final output. Then I'll turn off LC to get the best glossies possible.

[vlado]

Re: Light cache, glossies and fly-through

I've been using 1.4703 and learning about/testing light cache. It took me awhile to figure out that LC is stored in the IR map. I came across that here on the forum. I've noticed that when glossies are stored in LC, that a scene with lots of glossies will render faster with the LC in use rather than having secondary set to None.

This is because the light cache also stores the direct lighting, so there is no need to recalculate the lights for glossy rays. When you don't use the light cache for glossy rays, the lights must be evaluated in full with all the shadows.

I realize that Fly-through is meant for a single animated camera, but it apparently works for multiple cameras too?

Yes; in fact, you can make things so that your multiple cameras are just different positions of the same camera at different times - and in that case, you can use the fly-through mode to generate a single light cache for all the positions.

Best regards,
Vlado

[vlado]

Thanks wouter. I will try your suggestions. I have been doing a lot of research, here and with the online help (which I can no longer access since the site change).

You can access the help index here:
http://new.chaosgroup.com/support/private/2/
If you are having problems, please let us know about it.

Best regards,
Vlado

[czoog]

Thanks wouter. I will try your suggestions. I have been doing a lot of research, here and with the online help (which I can no longer access since the site change).

You can access the help index here:
http://new.chaosgroup.com/support/private/2/
If you are having problems, please let us know about it.

Best regards,
Vlado

I can't acess this area either. Where I try to login, it doesn't work, no errors or anything, it just takes me right back to the login page.

[siliconbauhaus]

not working for me either

[Frances]

Thanks Vlado. I found a nice little script by Michiel Quist ("cameranim01.ms" @ www.scriptspot.com) that converts all scene cameras into a single animated one. Came in handy.

[CCS]

A little bit of a tangent, but I want to throw something in. I love the look of vray and all the flexibility it provides. Just about any type of image can be produced. And I appreciate all of the features that continue to be added - especially in recent months.

But along with all this, my frustration has been growing as well. There are so many quirks and workarounds, it is hard to remember them all in order to get a project out the door. Most of it has to do with automation. For instance, when I reuse a stored IR calc, why are all of the settings still available? And when I reuse an IR calc that includes LC or QMC information, why isn't second bounce turned off? (or is it necessary "sometimes"? it's hard to keep it straight)

I realize some of this is in the help - but a lot of it is not. But just having it in the help is not quite enough IMO. There is too much that is just not intuitive. What might help is some sort of preset combinations of settings for various types of renders. For example, a method that would precalc lighting solutions, save them, then apply them to final render. Or an optional method that would turn off rendering, calc a solution at 1/2 size, adjust the size back, then turn render back on and run the final. Or presets for rendering for different purposes. (The animation presets are on the right track!) I realize many of these can be created and saved for reuse, or even scripted - but like materials, it would be great if there were presets. Then we could start from there and tweak as we need to. Right now everything feels like a tweak. Kind of like keeping my car running.

I keep feeling like most of the settings we're dealing with now should be under the "Advanced" tab. Things that I know about and can change if I need to, but don't have to weed through continually.

There are just so many peculiarities to remember, it is getting a bit overwhelming keeping track of them all. I still think vray is going in the right direction, and is one of the best all-around engines out there. It's just that this isn't a hobby for me anymore. It's day-in, day-out production. I would like for it to feel more predictable so I can get the next job out the door without trying to find that one magical setting.