Render setup details

Some quick details of my render setup while i'm rendering a scene out.

Once i had lit my environment to my satisfaction, of which the render times were running about 6 mintues a frame, i decided to do some tests with the characters in there. Adding the characters caused the per-frame render time to shoot up to about 11 minutes. This was unacceptable, so i decided to render the characters and environments separately. Since my camera didn't move in very many shots, this meant that i saved a lot of render time, as the characters only took roughly 2-3 minutes a frame.

There were two main challenges in doing this - light matching and shadows. I'll briefly cover the differences between my two lighting systems, to explain why i would need to light match. The environment lighting had 3 main types of light. Firstly, there were a number of photometric area lights used as the strip lights for the ceiling. These would provide a lot of fill light for the scene, and i also set them so that any camera that looked at the ceiling would see their area as white rectangles, like a real ceiling light. Using photometric lights instead of normal mental ray ones enabled me to use photometric light distribution lattices, which emulate the way lights shine out of their housing (in this case through a piece of textured plastic).

Next i had a setup of a skylight and some MR Sky Portals. I set up the skylight to have a gradient running from a light violet to a pastel orange. These lights give some vibrancy and colour to the area around the window, and make some contrast with the darker corners of the room. Thirdly, i made a MR Area Spot light, and put it very, very far away from the scene. This would simulate the direct sunlight shining into the classroom. I used an mr area spot instead of a basic direct light because i wanted to use Ray Traced Shadows, because of the softer edge it could give to the shadows. In the screenshot below you can see the lighting setup.

This lighting solution was great for my environment, but lots of photometric area lights combined with all of my characters' high resolution textures and hair was producing too long render times. Instead, i started with a basic 3 point lighting system, and then added a number of extra fill lights, to simulate the ceiling lighting, light spilling in from the windows and the global illumination used in the environment render.
 

 

Matching the light between the environment and characters was simply a case of fiddling with the light settings until i had one i was happy with, making sure the main light came from the same direction as in the scene, but also making sure i had nice detail on my characters. I could then comp this together in after effects. Below is an example of a shot i comped together.

 

Initially i rendered the characters in an empty scene, however this caused problems. When i comped the characters on top, they looked like they were floating on top of the scene, rather than sitting in it, even if the light matched convincingly. In order to solve this, i had to put the scene geometry back in, but set it all up with matte/shadow material, and re-render. This gave a much more convincing composition. Below you can see and example of a shot with the shadows included:

A further problem i had with this compositing approach was their hair, especially the little girl. When Max renders out the objects in the scene that are translucent it assigns the colour of the background to any pixels that are not completely transparent. I had initially set this up as a blue colour, to simulate the sky on any shots that had any sky visible. You can see the problem in the shot below:

 

The solution to this problem was to use the rendered shot from the background as an environment map. This would make any translucent parts already have the correctly coloured pixels behind, and make the characters fit in the scene better. The same solution ensured that the priest's reflective glasses would reflect the right colour (although the image would not be physically accurate, as the geometry did not exist). The shot below gives comparison. The saved stills would still have the alpha channel on however, as this would enable more control at the compositing stage.

A completed render of a frame from my animation.