This may be not quite the right forum to start what I hope is a discussion, but it seemed as good as any.
This is a general question about vray directed especially at those of you who are “rendering experts” : Whereas many non-GI renderers seem to be more or less as bad no matter what scale you use them at (whether you render a teapot or a car or a building), Vray seems _much_ better at small scales.
For instance, small product-design scale objects often appear 100% photoreal with relatively little work. Medium room-sized images (cars and interiors, mostly) take a lot more work but I’ve seen some instances where they look really quite close to the real thing.
When you get up to building-scale objects, it falls apart. They tend to look like scale models of real things, at best. This even holds true for rendering practices that do an incredible job at the large-scale-interior level, like Osmosis.
When you get to _really_ large objects (like skyscrapers in the city) things get a little better, but often you’re so far away its more just photoshop tricks anyway.
How can one fix this? Is it possible?
One of my colleagues here on the board had a theory that it was the GI light dome, and that if you made your models way way way smaller, then the effects of rendering would be more realistic. This seems unlikely to me.
Some other possible causes that have occurred to me:
1. not enough use of DOF
2. just not enough detail or realistic enough material use
3. poor integration with context or no context
4. not good enough entourage
I believe it’s pretty much a detail thing. A building has many subtle details that people usually don’t bother to model, either geometrically or as materials. This is why such scenes often look schematic and toyish. All surfaces have bumps, scratches, dirt and dust which help the obsever to infer the object scale. Camera parameters (position, fov) and even atmospheric effects (distance fog) all help to define the scale of a large structure.
Scaling the scene itself will not change anything in the lighting like exemplified on this page:
vlado, shouldn’t the indirect illumination automatically change, when the object gets scaled down or up?
i mean if you light out a small physical model of a building the light-distribution will look different than the same building in real scale?
as the model is much smaller, the light bounces much stronger and can light places which wouldn’t be lit when in real scale, or am i wrong here?
i remember that for lightscape it has always been said to model in scale to get a real light-distribution. is this not the case with vray?
Very important for my eyes is a natural lighting. Many users take a simple colored arealight in the windows, but I think it is not enough. The arealight in the window send to much light to the ceiling, the lighting is to homogen. A real world room with living lighting show color gradients at shadows and walls. In extrem cases (small windows or the window is closed by a cover at 99%) a camera obscura effect is visible. Here an example:
outside view of a house and the projection on the wall
The simple camera obscura effect makes, that the ceiling get a dark brownish or greenish color and the ground a blueish color from the heaven.
Here a schematic of the lighting:
What can we do? Use the area lightportals in the windows and a HDRI in the environment slot. This helps a lot. For exterior scene use big arealights formed to a cube around the scene, in portal mode again.
At the moment I have not ready to much perfect examples but here some WIPs from a current project:
If you scale the lights with it, or if you only use environment lights, then GI effects are scaled with the object to the exact same degree as the object itself; if you also scale the camera, you’ll get the exact same image as when the object+lights+camera were larger. This is because everything in the rendering equation deals with relative distances and angles, and not with absolute scales.
i mean if you light out a small physical model of a building the light-distribution will look different than the same building in real scale?
as the model is much smaller, the light bounces much stronger and can light places which wouldn’t be lit when in real scale, or am i wrong here?I’m afraid you are wrong. The way light bounces off a surface does not depend at all on how large that surface is. Like I said, if you scale the whole scene up and down (including lights and their intensities), you’ll get the exact same lighting result.
i remember that for lightscape it has always been said to model in scale to get a real light-distribution. is this not the case with vray?It is recommended to work with real units only because light intensities are typically strongly related to them (e.g. light intensity is may be specified in Watts, and this automatically “locks” the scale you must use). If you scale the scene and adjust the light intensities accordinly, even in Lightscape, you’ll get the same result.
A couple questions based on your lighting advice (which is much appreciated).
1. When you’re doing an indoor scene, do you tilt the window lights down a bit to reflect the sun, or do you add a second non-rectangular light (I would do this in max, with vray shadows on, directional), to mirror the effect of direct sunlight coming in (warmer of course) or do you just use the hdri for that? The scene you have is excellent, but overall it feels very much like an overcast day and I don’t know exactly in Rhino how to get the feeling of sunlight.
2. For exteriors, with your arealight/box setup, do you again use a directional light for the sun and shadows, or do you use the hdri, or… ?
3. For interiors with no natural light, would you slightly vary the light color of your light sources if you had multiple ones (see my theater interior in the gallery) or just give it up and use other cues to get realism? (this of course happens with real light fixtures as they age, degrade, etc).
1. When you’re doing an indoor scene, do you tilt the window lights down a bit to reflect the sun, or do you add a second non-rectangular light (I would do this in max, with vray shadows on, directional), to mirror the effect of direct sunlight coming in (warmer of course) or do you just use the hdri for that? The scene you have is excellent, but overall it feels very much like an overcast day and I don’t know exactly in Rhino how to get the feeling of sunlight.
I have not use any direct sun up to now. I have used the HDRI only. Important - if you use a light portal box around the scene you get a better GI calculation. The HDRI is a free HDRI, but I don’t have the link now here. If I find it, I send the link later. Here an other link: ぶっちゃけNIGHT – 暇だからガールズバーでバイトした時のことを語るね
2. For exteriors, with your arealight/box setup, do you again use a directional light for the sun and shadows, or do you use the hdri, or… ?
Only HDRI. But don’t use a HDRI with a strong local light like the sun. In my tests it has dosn’t works good. Better you use a directional light or … .
3. For interiors with no natural light, would you slightly vary the light color of your light sources if you had multiple ones (see my theater interior in the gallery) or just give it up and use other cues to get realism? (this of course happens with real light fixtures as they age, degrade, etc).
I would use different colors for different types of lights - bulbs orange, tubes greenish and outside lights blueish. But more variation I never have tested, but sounds interesting. In a room with many tubes you could build some groups and vary the colors.
Your camera view looks like a spy cam. Bring it down to a human level. And for a scene of this type, you definately want your vertical lines vertical with no paralax.
Also, the camera obscura effects you are laboring so hard to duplicate are entirely lost on a client. While I agree that the second render is far superior to the first, the first one is just a bad render all around and nobody with any sense of what is good and right would find it acceptable.
Adjust your perspective viewport as closely as you can by eye and then go into viewport properties and make sure the camera and target are at the same height. I think PS filters often do more harm than good.