Universal settings still universal?

[3ddesign]

Vlado's original post for "Universal Settings" was over four years ago so I wonder if they are still universal. Would someone please endorse or refute that post?

Also, I've been trying to find a good comparison between different GI engines but I'm still confused. I've been using Irradiance Map for primary and Brute Force for secondary for years. I primarily render interior still images using DR on a 16 computer render farm and the IR/BF settings I use produce great results and renders fast.

But more and more my consultants are using Light Cache and I frankly don't understand why. When I try to render with their models I don't like the outcome. Could someone point me to a good comparison between the two GI engines and what their benefits are? Maybe someone could try to do a sales job to convert me to Light Cache.

Thank you very much for your help.

[glorybound]

The only way I have been able to learn V-Ray was to actually test everything out. I would test those settings against your typical settings. I know, this comes from a guy that has asked every question in the book

[MoonDoggie]

Bookmark this: http://www.spot3d.com/vray/help/150SP1/

Here's what is pertinent to you: http://www.spot3d.com/vray/help/150S...s_interior.htm

[vlado]

The universal settings are still valid (with one exception - when you do an ambient occlusion pass through VRayDirt and VRayExtraTex render element you should *not* set the adaptive amount to 1.0; I missed this possibility when writing the original post).

However if you are asking for comparisons between the different GI engines, then this is a different topic. The light cache gives you more a accurate lighting distribution as it can handle multiple light bounces. Brute force for secondary bounces will typically give you darker and more contrasty GI. If you don't like the light cache results, then it is likely that your materials have relatively high diffuse reflectance (i.e. pure white or close to pure white) causing renders that are washed out. (Not sure if you are using it, but the whole "linear workflow" stuff usually helps with this.)

Best regards,
Vlado

[joconnell]

I've always been curious about Light cache in that regard - it's tracing rays backwards from the camera, recording each surfaces colour and reflection levels to measure light loss, until it finds a light source to begin from. As you mention on spot 3d this is the opposite to photons which fires thousands of samples with the hope that they'll bounce around the scene usefully and contribute to surfaces seen by the camera in the render. What happens to samples that don't ever get to a light source or those that leave the scene off into the empty space outside it?

Also with the IRmap, since it's being done from the camera rays and also the direct light source, there'll be a large portion of it that doesn't immediately point back into a light. Is it a case where again vray works out the energy loss from each surface and then feeds in the rest of the lighting sum from the lcmap to give you a complete path? I take it that this bounce is pretty much just substituted for the first bounce of the lcmap. Effectively the lcmap is a complete lighting solution, it's just that it's poor quality for small details which mainly only occur in the primary bounce.

Cheers!

John

[3ddesign]

Thanks for the replies. I was always under the impression that IR/BF provided a more accurate solution but since my post, I've been testing LC and the universal settings and am quite pleased with the results. I especially like the way the shadows looked. But the renders still take longer than using IR/BF due to being able to DR the GI calcs. I read somewhere that LC would be DR able some day. Is this true? This is critical for me because I'm often asked to complete a render in a matter of hours. Also, Vlado mentioned in the universal settings thread that it might be possible to tweak the settings to increase speed. Can someone point me to a thread that shows optimized LC settings for interiors? (Although from the searching I've done, it appears there are as many optimized settings as people using them.)

Craig

[joconnell]

The nicer shadowing provided by Light cache is possibly down to the fact that light cache lets the light bounce around your scene almost infinitely until it dies off enough to no longer make a difference. Brute force has a set amount of times it allows the light to bounce set by yourself. It uses 3 bounces by default and you should find that upping the bounces will give you a brighter image due to more light being added to the scene. You might also find that adding in another bounce of light only adds in a tiny bit more brightness to your scene but adds in a huge amount of extra render time.

In terms of brute force vs Light cache, brute force works as its name suggests by calculating GI for every single pixel in the image. As a result of this you get lots of tiny detail in the image, but having this much detail can also make your image a tiny bit grainy. Personally I like the look of grain since it can be quite filmic but too much grain is a bad thing. Light cache on the other hand fires much fewer samples but lets them bounce until they don't makes much difference anymore. When all of the samples come back, they look a bit blotchy like this:



and vray simply blurs and filters them to get a smooth result. The big difference is that with GI, all of the really nice sharp detail is in the very first bounce of GI so that's where you want a really high quality method like the IRmap or brute force. From the second bounce onwards, the light gets very soft so you can get away with using a much less accurate method but still have really good results.

In terms of optimized settings, I think that LC is by default very quick and what Vlado is talking about is more to do with overall sampling and anti aliasing. for example you might have a scene with really large amounts of motion blur and glossy reflections that his universal settings will be really quick for. Then again, you might also have another scene that's really simple and doesn't need the high anti aliasing values that Vlado's method uses so you could use much lower settings but still get a clean result. Every scene is always different so you might need more less AA or more or less IRmap samples. There'll never be one set of settings that are the best for every scene out there. Vlado's method is more something that'll pretty much always give you a really clean image.

[3ddesign]

Thanks. That explanation helps.

Do you know if LC will ever be able to be calculated over DR? Or maybe it is fast enough that no one cares. I guess I just like seeing all those buckets floating around my screen.

[BLADE]

One LC over DR would be nice since now every DR slave is rendering it´s own LC which not used in the end.

But LC is rendering in passes per CPU. The more passes you have to blend in the end. the more smoothed is the result. So if you are doing it with 16 PC over DR (and every PC with quadcore CPU) you are ending with 64 passes. From what I have understand, from smoothing this 64 passes you will loose a lot of detail.

Regards
BLADE

[simmsimaging]

I tend to keep the LC subdivisions at 800-1000 for hi-res stuff. I don't do interiors, mostly product type things, but I find the detail is pretty good at that level, and renders pretty fast even at hi-res (6k-8K is typical for me). It's rarely the LC that feels slow in the process to me. That said, check into SolidRocks - I haven't played with that part of it, but Jerome has something in there for sharing the LC across the network - it might do just what you want.

b

[joconnell]

In my experience LC takes a few seconds unless there's something wrong in your scene.

[3ddesign]

Then I must be doing something wrong. Here are two images based on the Flipside scene with a plain Vray mtl and three vray lights. The render using BF/LC and Vlado's universal settings is grainy and took 8 minutes to render. The render using IR/BF is a bit darker but has less grain and rendered in 28 seconds. Both were rendered using DR over 5 quads.

[vlado]

It would be even faster if you use irradiance map/light cache, which will give you the same result in terms of brightness.

Best regards,
Vlado

[joconnell]

Nope you're totally spot on.

Brute force is by far the most accurate solution in terms of capturing small details since it does GI calculations on every snigle pixel in the render. Because of this though, it comes back with so much detail to deal with, it takes vray a while to clean it up with it's anti aliasing. If you use brute force, especially for interiors you can either use loads more brute force samples which will give you a cleaner results, or take a noisy solution and try to use anti aliasing to clean it up.

Irradiance mapping on the other hand starts off with a low res GI calculation of the frame, looks at the areas that have more detail and refines those areas only. You end up with a softer solution than with brute force, and because of that vray can use less anti aliasing since there's less grain to take out.

Being totally honest I'd never use brute force in an interior first of all if it's only a still or if the only moving thing is the camera. Irmap and LC is a far quicker rendering solution in these cases - it's not quite as sharp as using brute force + LC but the render time is so much faster it's hard to ignore.

[joconnell]

Vlado - could I send you a quick mail with a few questions on motion blur and gi calculation? I'm going to write a few explanations on things but I want to make sure my understanding is correct to start with!