Methods and Techniques for Lighting Hair: Area Lights
Author: Kim Aldis (http://www.aldis.org.uk)
I’ve been playing with my hair a lot these days. There are some days when I actually feel like pulling it out. Mostly, I’ve been determining the optimum lighting choices for hair for the current project I’m working on. I examined all the light types, including the use of soft shadows, and have come up with these guidelines for getting the most from standard XSI light features. It’s cheaper than a trip to the barber. Let me know what you think and feel free to record your own “tips” in the comments.
Today, I deal with area lights.
It goes without saying that area lights are expensive and should only be used when all else fails. The appeal of area lights is twofold: You get soft shadows because the umbra size varies with distance between shadow casting object and shadow receiver and you get a desirable softness in the overall lighting.
You’d expect that the degree of softness would be the most important factor governing render times but as the tests below show, this may not be the case. Increasing the area size from 1 to 5 only slightly increases the time to render. It’s the simple act of turning on an area light that bumps the render time from 6 secs to 1:30 or thereabouts.
 |
|
 |
|
 |
|
An area light's sample count controls the number of scattered samples cast from the light and is a considerable factor in rendering times. The images above were rendered with 1 light, sample size 6. Increasing this to 12 resulted in a render time of 2:51, nearly double. Typically, you increase sample count to reduce any graininess in the shadow umbra.
The size of the umbra is dictated not only by the size of the area light but also by a combination of the distances between the light, the casting object and the receiving object, as illustrated below. Note that infinite area lights cast shadows without umbra since an infinite light is an infinite distance away from any object.
Area Lights and Hair
Rendering hair with area lights changes the ball game a touch and the render time difference between area lights with and without shadows is considerably more exaggerated, as indicated in the render times of the images below: a whopping 44 minutes with shadows compared to just under 2 without. There is no motion blur in these images.
 |
|
 |
|
Note the coloured shadows in the second to last image, caused by the transparency used in the hair shader. A workaround to this is to plug the shader through a colour correction node with the saturation turned down before it goes into the hair’s shadow input.
Tomorrow I'll post my notes on point lights, spot lights, shadow maps and more.
