Helldoor's Test Field | Interior

[Helldoor]

"There are two secrets to success: 1. Never tell everything you know."

Well, screw that.

Some yars ago, after finishing my Masters in architecture, I decided to work on my own in the 3D field. Rhino was the software we used during the studies, so let it also be the one for commercial use. Paired with V-Ray, of course.
But damn. Although the combination of these two rocks, there's also almost no how-to database about rendering using them (compare with Max +V-Ray). Good old 'trial and error' was the order of the day.
And I guess a lot of us in this part of the forum may have had similar experiences on their way.

This prelude leads to the core of this thread - putting some know-how together in order to learn from eachother and achieve better renderings, directly out of V-Ray. I don't know if anyone would be eager to share his hardly earned experince and knowledge, but I hope it. I'm not the best, but I'll share everything I know.

So, let's go. Recently I built the minimalist and very warm interior of Bernardo Bader's Moor House (Haus am Moor). The purpose - having an interior surrounding for my 3D Objects (furniture, deco, lights,...) and playing around with the new V-Ray. Later, I'll finish the facades of the house and also make an exterior test field with the same purpose.
Now, about the V-Ray part. The renderings come out decent. But they could be a lot better.

| Living Room / Kitchen [1600x1130 / no PostProd / Render time: 48 min]

The Environment Lighting is not 'warm' and realistic enough, it also 'burns' the exterior parts you get to see from the inside.
Also the carpet doesn't look good at all. I wanted a furry one, using a fur object, but the single strings of the fur object seemed transparent, and didn't cast or receive shadows??
Overall, it could be more realistic.

Heree is the setup I use so far:

_ Render Settings

• Renderer | Custom (based on High Quality) - because I lowered the Raytrace Noise Limit manually;
  Raytrace | Noise Limit 0,002;
  GI | Irradiance Map (-3,-1) 70 Subdivs 40 Interpolation + Light Cache 1500 Subdivs;

_ Lighting Setup

• Window light | GI (Skylight) (Environment > Environment Override) - HDRI + x10 multiplier;
  Indirect Light (kitchen) | simple Rectangular Light;

_ Scene Setup

• Measures | Centimeter - Good for big structures, but still accurate enough for small stuff;
  Fragmenting | For a clearly readable structure I'd let small gaps between two objects/faces (e.g. a chair 'floats' two milimeters above the floor;
  Material Usage | IMO, using a lot of different materials makes the scene appear unhomogenous; if I have to, I'd firtsly go to PS and try mathing the tonality of the different textures as much as possible;

_ Material Setup

• Color Correction | I'd gladly use Color Correction in the Diffuse slot on some materials to be able to quickly adjust them without PS;
  Displacement | For the most materials I'd use a Bump MAp, but for some - the Paruqett or a Carpet. I'd use displacement;

My wish is to hear and try your out ideas, updating the scene set up, comparing and testing different approaches until - in the end - we have a good, set up for interiors. A V-Ray settings file to download + a bunch of tipps about interiors.
If this succeeds, I'd make a second thread about exteriors.

Thanks to all of you, who want to praticipate in advance.

[Helldoor]

This is what the living room image looks like with literally 10 minutes of PS:



But I'm sure, that same or at least similar outcome could be reached using only V-Ray.

[Micha]

Nice model and good state. Nice idea to create a scene for experiments.

The floor and the table wood could be more reflective and maybe you need a higher texture resolution, the bump gaps are looking to big.

Did you tried noise threshold 0.01 or 0.02? I use 0.015 daily, for high res 0.02.

You could turn on the lights at the table.

Did you enable GI AO? Helps to bring small contact shadows. Intensity 0.8, radius 10 and 32 subdivs could be ok.

"use for glossy rays" at the LC is turned on?

The windows needs more details like frame and opener.

A sun light on the floor can be a nice addition, also you get some warm color bleeding to the room.

Materials - I would try to add some additional materials like glas (maybe a glas sheet on the couch table) and green plants.

And maybe you find a high res scan 3d people, some sides offer free ones (renderpeople.com, axyz-design, ...). Maybe someone is looking out the window.

You could add some images on the walls - poster or so. Here you could bring in the glas material with reflection of the room.

For me the greenish color bleeding from the exterior is ok, it break the warm mood.

Enough for today, have fun.

PS: ... later create a VR pano.

[Helldoor]

Thx, Micha! That's some wonderful input.

Before I come to the single points, I should maybe post the original interior pics - these would make some points easier to talk about:



_ The floor and table are pretty diffuse and bright in the original. I went for the oroginal look, but I'm not really happy with the parquette either. Maybe I'll try some smaller mapping (as I like the texture) or go for a higher res texture.
_ Current Noise Limit (the new Noise Threshold?) is set to 0,002
_ Turning on the lights will surely be cool. The only uncertainty I'd have about it, would be washing out the bright table textrue. But I'll try it. What kind of light would you use - I'd go with a point or sphere one?
_ Ambient Occlusion is enabled - Amount 0,65 / Radius 10 cm; using the Progressive Renderer Mode I cannot find where to set the subdivs
_ Same goes for the 'use glossy rays command' in the Light Cache - I don't find this option anymore (btw, years ago I learned always checking this option from you)
_ The windows that can be seen, are the 'built-in' ones - they simply don't have frames and openers (check interior fotos)
_ Would you use a V-Ray Sunlight or Directional Light for the Rays on the floor? Also - wouldn't the sunrays bring a lot of contrast and make the dark areas even darker? Whatever, I'll just try it.
_ Bringing some reflections/glass on the walls sounds awsome - I can imagine some really nice framed SW poster on these walls.
_ What do you mean about the VR panorama?

Update will follow soon. Btw - right now I'm using Lanczos, but I know Catmull Rom is often chosen for architectural purposes?

[Micha]

_ you could multiply the floor texture by 0.7, so it will be not so bright in the lighted areas
_ I would be curious for a 0.01 noise test
_ try warm spot for lighting the table and add an emitter bulp inside
_ "use for glossy rays" - you seems to use VfR3, than this option should be on behind the scene
_ never I use the sun light in the past, to much trouble. I was glad with HDRI env and dir light (shadow radius 0.03)
_ VR pano - I thought on some like a simple pano that can be created for example per pano2VR

Since you seems to use VfR3 you could make an engine test. I would be very curious to see your times and quality if you render LC+IM and LC+BF. My impression is that using BF is slower, but I would like to see I'm wrong. Chaosgroup recommend BF, but I'm not sure it's realistic for tight deadlines and many renderings.

_ my impression is that the light transport is to strong, dark areas get to much indirect light. For example at the last image the area under the wall couch.
In the real world bright white has a diffuse light reflectivity of max 80%. But Vray white is 100% indirect light reflecting. At VfR I set the secondary engine to 0.8 for this reason. At VfR3 the option is missing. (I'm to lazy to ask for and to get no answer now. ) Physical correct would be to set a multiplier 0.8 for all textures and colors. But if you do it for the bright main materials only you could get 90% of the effect. So, you could try to set your bright wood of floor and walls at 0.8.So also the pyramid ceiling could show more light attenuation from down to top edge. At the moment it looks extreme homogeneous lighted.

[Helldoor]

Once again, big thanks for the input, Micha.

_You guessed right - I'm using the V-Ray 3.4 beta. And I don't know if I'm currently too retarded for it, but I don't find material multiplier among the material settings...
_ I'll check the noise levels and corresponding render times at 0,002 and 0,01
_ for the table lamps I'll check point light vs spot+emitter
_ so you yxou mean in VR3 'use glossy rays for..' is always on?
_ one more test - HDRI+Sun vs HDRI+Direct Light
_ and one last one - the engine test; I always use IMap as primary and LC for secondary; you suggest LC (primary) + IMap (seconday) vs LC (primary) + BruteForce (secondary) - do you mind explaining, why you'd test that way;

There's a lot of testing ahead. I'll post, as soon as I have some outcome out of it.

[Micha]

_ I checked it here - right, the texture multiplier is lost (bug)
_ Fer told that "use for glossy rays" is on for LC always
_ use LC as secondary always, I didn't type it in the right order. BF is a computing without interpolation, it's 100% correct, without big shortcuts in the calculation. IM is the fast way with a lot shortcuts and interpolation. I use IM all days. In the past we was able to mix IM and BF per detail enhancement, where needed, at small details, BF was used. Feature is canceled now. Chaosgroup says, BF is better than before and can be used all days. In my tests it was to slow for my need. I can't wait several times longer for an image. I'm on the time limit, since I render a lot of interiors at 10K resolution.

[Micha]

I found the texture multiplier. It's hidden in an own multiplier section at the end of the material options. I hope we will get it back at the placed where most needed, but I'm not sure the team is reading the bug reports anymore.

[Helldoor]

Hi, thanks, I found it too. Hmm, don't know what the majority says - maybe the most users like it, when all the Multipliers are grouped together?
Btw - where can I report bugs - I noticed, that copying a material and changing the set up of the copied one, also changes the setup of the default material - as if these were linked...

And now some recent test results so far:

_ Render Engine Combination Test | Brute Force+LightCache vs Irradiance Map+Light Cache vs Irradiance Map+Brute Force [Renderer: High / Noise: 0,01 / Render Output: 1600x1130]

Well, the first posted picture should've been the outcome of Brute Force (primary) + Light Cache (secondary). After two hours of rendering however, the image reached only just 5%, and I broke it off. If I find a way to render using BF in primary, but also having decent render times, I'll post the pic here. Until then, it would only be the other two:

| Irradiance Map (primary) + Light Cache (secondary) [Subdivs: 1500 / Sample: 0,02 / Retrace: 4] : 00:59 h
| Irradiance Map (primary) + Brute Force (secondary) [GI-Depth: 3] : 01:32 h

Both images are almost 100% identical. After switching the image preview back and forth several times, you'd probably see, that the LC image hat a bit brighter and warmer tone for the wall and curtain textures, which I personaly like more. And then comes the render time comparison. Brute force = brute half an hour more. And that's the restult for a small resoultion such as 1600x1130. That'd be exponentially (a lot) more for a 4k rendering, for example (or for Micha's 10K). So, I'll be using Light Cache further on. BUT! Using Brute Force could be very useful for a fast, but still pretty clear render image preview - the engine gives a quick and very detailed preview of what the final image is about to look like in the first minutes of rendering.
Later today, I'll add up here some outcome of the Noise Limit test.
As always - any further advice and ideas for creating a nice and realistic V-Ray SetUp for Interiors are welcome.

[Micha]

I'm not sure I right understand - the light cache should be the secondary engine in both case because LC is the big advantage of Vray. The LC is calculating the general raw GI lighting in a first pass. The primary engine is used for the visible solution (Sorry, I can't explain it.) We new recommendation is using BF as primary engine. I afraid it's a lot slower than IM. Did you use the IM in your "LC" version?

[Helldoor]

I used IM as primary for both rends. For ther first pic I took LC in secondary and for the second one - BF in secondary. I'll render once more with BF in primary and LC in secondary and update in the post above.

[Micha]

OK, so we have the reason for the dark curtain. You don't used the infinite GI bounces of LC, only the 3 bounces of the BF. Now the basic light solution should be the same in both tests and now the difference should be like little details are handled. I'm curious.

[Helldoor]

OK, so we have the reason for the dark curtain. You don't used the infinite GI bounces of LC, only the 3 bounces of the BF. [..]

Well, that sounds pretty logical. What I don't get is, why LC takes less time to render, although it has a lot more bounces than the 'only' 3 from BF.

[Micha]

Sorry, I don't know so much about theory in detail.

https://docs.chaosgroup.com/display/...Light+Cache+GI

[Helldoor]

Well, I guess Brute Force (primary) + Light Cache (secondary) won't be an option for me. I started rendering two hours ago and now it has reached 5% (Noise threshold lowered to 0.034483 with 75.3 percent active pixels.)! Don't know if I'm doing anything wrong, but this cannot/shouldn'T be. I'll stop it and wait, if eventually someone knows any better soultion to rendering with BF in primary.