Tips for more light through the window into the interior Tweet

[tomas.base]

If anyone has other tips for interiors and want to share, I'd be interested.

Sorry, I just scribbled it into one picture
A somewhat experienced user will know. It probably won't help beginners much, you need to know the functions a little, what they do. And the basis of everything for me is CHROME "filters". I like it when the output behaves more like a camera and I try to adjust the IPR preview so that there is not much post-production.

I only use this procedure when I can't get light into the room and it's good if you don't want to add light (fake-lights) or if you don't want to have more windows (holes in the wall) out of the shot. Or you simply don't want to use interior lighting. Such images do not have nice soft shadows. In my opinion, such light looks better than area lights instead of window glasses.

Gradually increase the diffuse samples until there is no difference in the lightness of the walls. A full scene is better than an empty one, where the light doesn't reflect very well everywhere. 32(64) dif / 24 spec is enough for bruteforce, the brightness is not better, but it looks more natural. Huge specular numbers are only good for glass stuff so there are no black spots (render easily hundreds of samples over the region).

Otherwise, I will render the final render without window glasses, and I will make the second render as a window region with glass. The glass blocks the intensity a bit and sometimes "dusts" too much into the noise (triple glazing). But if you leave the render for several hours, the result is more natural to reality. I also render simpler scenes with glass, it takes longer, but the light diffusion is softer.

I only render ORBX exports in Standalone, because the plugins can't save the render state (there's nothing worse when the PC crashes after 10 hours of rendering) and unfortunately for me Lightwave doesn't have the new 2022 features.

But post-production is a know-how that I won't publish, because the process reflects my imprint and every graphic artist has his own, and I don't think I'm good at post-production, it's often trial and error.

Standalone is clearly the best, most stable and fastest for rendering, including the GUI. I only use the plugin to build the scene, because in Standalone, I don't know why, it is not possible to move anything in realtime as comfortably and quickly as in the extra editor.

I don't remember Standalone ever crashing unless the GPUs overheat. I'm using Precision X1 (2x RTX 2080Ti) to reduce performance based on ambient temperature.

I wanted to give something back to the community, maybe it will help someone, sometimes I add materials to LiveDB.

Maybe I'll get back to you with other tips in a year or two I probably don't have much more to say about interiors

Tomas
I work as a contractor and most often create visualizations for various industrial/life-style products embedded in interiors.


note: translated via Goole translator, but the AI ​​is already quite good. Grammar is not my strong point

[tomas.base]

I forgot the most important thing.

The most important parameter when using the DayLight model is parameter TURBIDITY.

Larger value, more light scatter. I don't know how to describe it, but something like volume, but without fog.

It works correctly and logically only in the Hosek-Wilkie model and is best for the interior.

Instead of black ground, put a reflector (invisible to the camera or below the shot level when the glass is in the window)

Without a reflector, the ceiling above the window is not so bright.

It is often enough to use the reflector as a sidewalk or road in front of the windows.

Turbidity:

0 - space vacuum, sharpest shadow
2-3 sun-day, contrast, blue shadows from the north, yellow light from the sun, hard and dark shadows
4-6 less blue from the ambient and yellow from the sun, not so hard shadows
7-8 blue and yellow (from the north and from the south) no longer colors the reflection, like a haze effect without fog, brighter image, less contrast of the light-shadow gradient
9-10 blue and yellow no longer have any effect, large and colorless light scattering, sometimes it is good to tighten the gamma

A larger number gives a smaller gradient between the darkest and lightest part of the space, the light gets behind the object and into the shadow, the shadows are very soft.

A value of 10 is the best solution for few windows, small windows, the image is completely free of color diffusion, shadows are soft, little contrast between light and shadow

If the turbidity value is high and the Sun intensity is reduced, for example to 0.7 - 0.5 - 0.1 and sunsize to 10-100-1000, you will get a strong haze effect without fog.

All values ​​preserve the light direction, the north side is darker and the south is lighter - this still affects the glare in the backlight.

For high turbidity values, it is possible to add contrast by reducing the gamma, to 1.8 / 1.5 and below, then the black spots typically under the sofa.

Turbidity does not work, if you want to use HDR for the environment in the DayLight model instead of the gradient (ambient from HDR, sun from DayLight), you must turn off the no-importance sampling option in the HDR tab of the DayLigt model, otherwise you will have two suns. It is a good idea to first align both sources (DayLight sun and HDR sun) so that north and south are the same for both sources. Since turbidity does not work here, the blue components have an effect according to HDR.

If the Sun intensity is 0, then the light is guided only by the gradient. A practical example is that Sun intensity=0 and Turbidity= 10. You will get pure white light and strong scattering, where the light reaches beyond the object it hits. It's almost everywhere. The shadows are the softest you can get. Even without the sun, the direction of the rays is preserved and you only have to position the sun as if it were shining, it simulates a heavy haze in front of the sun even if the ambient seems uniform, so there is still a darker north side and a lighter south side for reflections and backlighting. The picture will rustle for a long time. I strongly recommend using the portals, which make the most sense in high-quality ambient light (HDR with cloudy weather) or with a gradient (Sun intensity 0), but here the south side must be fully used, from which more rays come. If the portal is used incorrectly, it will not work. If it works correctly, the number of samples for smoothing is much smaller, maybe 10x. Portals do not like sharp light from HDR and high Sun intensity. If it works, rendering is slow, but at 1000 samples you're done, no portal 10000. Geometry must be closed, no window glasses. One 4 vertex ploygon in each window hole. Normals direction inside. Blinds and curtains complicate it, but if they are not in every window it is still possible. If they are not in every window or across the entire window area, smoothing is still better than without a portal. This applies mainly to very small openings. For example, you won't render igloos from the inside with one hole without a portal. Try a 2x2m cube with a 30x30cm opening for a portal and put the camera inside. Give only white color to Enviroment. You will see the brutal difference and how the portal works. Then you can tell if it works for complex scenes. The noise is specific with much less grain.

Hosek-Wilkie is the best prop light model for the interior, where you can influence the gradient of the Dailight model quite correctly. For other models it doesn't work or the algorithm is unrealistic and doesn't make sense to me based on real light on Earth.

[elsksa]

Hi Tomas, such post is kind of you to make. It raised my interest on some points:

I like it when the output behaves more like a camera and I try to adjust the IPR preview so that there is not much post-production.

These options shown in the Camera Imager for "creative looks" do not allow proper post-production work. Only EXR does, or log-encoded TIFF. Some more information here and here.

Gradually increase the diffuse samples until there is no difference in the lightness of the walls. A full scene is better than an empty one, where the light doesn't reflect very well everywhere. 32(64) dif / 24 spec is enough for bruteforce

By Diffuse Samples, did you mean Diffuse Ray Depth? There is not a per-lobe sampling as the written in the original text.
32 or 24 respective ray depths is very high for average interiors.
Some more information here about ray depth and GI Clamping (which was set quite low in the screenshot).

Otherwise, I will render the final render without window glasses, and I will make the second render as a window region with glass. The glass blocks the intensity a bit and sometimes "dusts" too much into the noise (triple glazing). But if you leave the render for several hours, the result is more natural to reality. I also render simpler scenes with glass, it takes longer, but the light diffusion is softer.

Have you considered the newly introduced Photon Tracing Kernel?

Standalone is clearly the best, most stable and fastest for rendering, including the GUI. I only use the plugin to build the scene, because in Standalone, I don't know why, it is not possible to move anything in realtime as comfortably and quickly as in the extra editor.

Standalone is a joy to use. Regarding your "issue", my tips with working in Standalone is to lower the resolution while working, and enable sub-sampling. Thanks to its node based workflow, there can be various render target node setups. One for low resolution and low kernel settings, an other one with higher resolution and kernel settings for test output or master.
I've demonstrated it here.

[Builtdown]

Thanks for the tips!

Would someone wanna share tips of outside lightning? I have some problems with outside shadows being too dark. Playing with gamma helps a bit but some areas go too light after that...

[elsksa]

Thanks for the tips!
Would someone wanna share tips of outside lightning? I have some problems with outside shadows being too dark. Playing with gamma helps a bit but some areas go too light after that...

Hi, first of all, never ever change what's commonly named "Gamma" in any renderer and in post. Simply said, it is solely a technical option, not meant to be used for aesthetic and creative look creations.

A short list of primary recommendations:
• boosting the lighting aka providing "the 3D scene" enough "light energy", so to speak. This is a must.
• not "abusing" GI Clamping with too a low value, as misleadingly suggested on many Octane content.
• using a post-rendering imagery solution such as this one. Actually, this is the only recommended one as of now.

I've shared this Octane Render Settings Guide among other informative content such as this one on Octane Lighting.

Do not hesitate to create your own forum post (thread) if needed.

[Lewis]

Great Tips guys, thanks for posting.

[tomas.base]

Elsksa, thanks a lot for your comments. I have read all the links and will try to install and set up OCIO as you recommend. I'm still an amateur

I understand that tonemapping is not ideal. I tried 32-bit linear, but Photoshop still can't work fully in 32-bit. I solve this by using a saved render-state for exposure/tonemapping changes, then converting the new EXR to 16-bit in Photoshop.

Many thanks and respect for what you publish on the web! I probably understood a few things better.

By Diffuse Samples, did you mean Diffuse Ray Depth? There is not a per-lobe sampling as the written in the original text.
32 or 24 respective ray depths is very high for average interiors.

Sorry, I don't understand very well the mathematics of the diffusion model in OCT. I just notice that more samples slightly lighten the places further in the room. In practice, I increase the values as long as it is reflected in the lightness. For example, if there is a plant with many leaves in the room or a fabric folded over itself, the transmission effect is better for larger diffuse values. It probably also depends a lot on how well the materials are set. Setting the specularity, glossiness, roughness and IOR parameters well so that they do not contradict each other is the most important thing. The upper limit of the GI is sometimes unnecessarily large and sometimes I prefer to set a compromise due to the noise and also depending on where the camera is looking, what angle the light has from the exterior. Many times I forget to work and tune several parameters for hours.

Have you considered the newly introduced Photon Tracing Kernel?

I have already tested the Photon tracing kernel, but only on glass objects that I render over the limit region.
But Caustic works very well, for example, at the pool, just to show:
pool caustics (hevc-avi 1.6MB 3s)
Simple water (no volume scattering) is procedurally generated by 3xRipples (additive blending) in LW as Deform-DispMap on sub-div-plane), anim exported to ORBX. Frame time about 10s, denoised (2xRTX2080).

Tomas

[elsksa]

Elsksa, thanks a lot for your comments. I have read all the links and will try to install and set up OCIO as you recommend. I'm still an amateur

We are all on a never-ending learning journey.

I understand that tonemapping is not ideal. I tried 32-bit linear, but Photoshop still can't work fully in 32-bit. I solve this by using a saved render-state for exposure/tonemapping changes, then converting the new EXR to 16-bit in Photoshop.

Photoshop is to eject from the equation. Enabling the "Tonemapping" checkbox is ensuring a limited post flexibility and excluding a master archive (rendered output) of the renderer, similar to photographing in JPG instead of raw. A 16-bit EXR (floating point) ≠ 16-bit TIFF (Integer). EXR @ 16-bit is in fact more than enough (more than any raw file from any digital camera sensor at any price, even +60k), just to put things into perspective.

Many thanks and respect for what you publish on the web! I probably understood a few things better.

Thank you for reading and the kind words. Do not hesitate to reach out for further elaboration or clarification. I will gladly help beyond my pages and improve them accordingly.

Sorry, I don't understand very well the mathematics of the diffusion model in OCT. I just notice that more samples slightly lighten the places further in the room. In practice, I increase the values as long as it is reflected in the lightness. For example, if there is a plant with many leaves in the room or a fabric folded over itself, the transmission effect is better for larger diffuse values. It probably also depends a lot on how well the materials are set. Setting the specularity, glossiness, roughness and IOR parameters well so that they do not contradict each other is the most important thing. The upper limit of the GI is sometimes unnecessarily large and sometimes I prefer to set a compromise due to the noise and also depending on where the camera is looking, what angle the light has from the exterior. Many times I forget to work and tune several parameters for hours.

Not samples. There is only one "master" sample slider in Octane. Ray Depth is simpler than it looks, math-free. I will spare you phrases and show an explicit demonstration: