[Asmithey]

Hello,

Not sure if this already an option. I have not found it yet. I do a lot of large scale site renderings. Sometimes I have to go set up cameras to do eye level shots within the large scene file. Does Maxwell only calculate what is visible to the camera. If not, is this an option that could be implemented to save resources?

Thanks

[JDHill]

Rendering itself is exactly that -- calculating what is visible to the camera -- put a few big mirrors both in front of, and behind the camera, and then try to define what it means for things to be visible to the camera. It takes a human to make decisions like that; in other words, if you know there are things in the scene you don't care about, you always have the option of hiding them.

[luis.hijarrubia]

Anyway, a very distant object can reflect light and add a slightly indirect reflection to an object that is visible on camera. That little indirect reflection is what really does maxwell renders so real.

Once some researchers tried to mimic a famous lunar photography from the Apollo 11 mission. They set the exactly camera parameters, sun on that day, materials… and they found a very similar result, but no perfect. The problem was they forgot to put on scene the astronaut behind the camera, and his helmet was reflecting enough light to have that discrepancy, when they fixed that the rendered result was almost equal to the real one.

[Asmithey]

There are some rendering programs, I think Form-Z used to do this, where you can use a volume to define the area/ objects that the computer would use to calculate the rendered scene. I think it would be a good option for quick results. By quick I mean saving time hiding layers and objects.

I will share an example scenario in a bit.

Here is an example. I have this large aerial rendering. It has four homes. Lots of tree models and one large site model. The initial contract was to only do the aerial. After I was finished, the client came back and wanted to do eye level renderings of each house. So when I originally set up the project, it was not managed with setting up the scene to do individual renderings.

If I would have know this I would have divided the whole thing up accordingly. So, it would be great if I could drop in a wire frame volume and just have Maxwell calculate only what is in the box. Only what is in the box would influence the rendering.

I could have one volume over each house.

[Mihai]

Which resources take the most RAM in your scenes, when rendering? Geo, textures, render resolution?

If the geo/displacement fits in RAM, it doesn't really matter. With instancing, I don't see how one can run into RAM issues even with very large scenes. "Large" in the sense of dimension, doesn't really apply in our virtual world. I can have a 1m tiny scene that can make my 16GB of RAM explode. I would think it's the host app that can run into issues regarding resources and things to keep track of, display them in the viewport etc. before it starts bothering the renderer.

EDIT: After seeing your example, you can imagine running into the issue of seeing just empty space for an eye level render because you choose to hide anything outside that box. And the lighting would look pretty "empty". How would you solve that?

I think I would do a 360 render of that forest, and use it as HDR lighting, and setting up a very simple second scene where you drop in each house/lawn etc. and then use the HDR for lighting and to not have empty space in the background. If you already have a background image, I would still think it's useful to have a 360 view rendered of that environment, for good lighting. Otherwise, you risk getting a render that looks like a house in the middle of nowhere - even if you do composite it over a background.

[Asmithey]

Am thinking of rendering time. I have plenty of ram and I am instancing all trees and plants. I think it would render faster if I only hade to render a part of the overall scene, correct?

To solve the blank space. I have backdrops that would fill that in.

The 360 idea is good, Thanks..