Quick Lighting Tip: How to Properly Overexpose Objects in 3D
Learn this simple trick to make sure that your renders can properly become overexposed when harshly lit within your scene.
I recently spent some time watching the video What 3D Artists need to know about CG Math and I came away with a new appreciation for lighting in 3D.
In this video, Kenneth Turner begins by talking about light rays in general, specifically scattered light (diffuse) and reflected light (specular). By combining the two, you are able to generate a realistic final render.
But then he goes a step further to introduce you to the simple math behind the results. The biggest takeaway being that any underlying math that results in a zero, will have negative consequences on your final render.
The left image demonstrates proper overexposure, and the right image shows what happens if your RGB values include a zero.
If you look at the base color RGB on the right image you would see: R = 0.8, G = 0, B = 0.
In the proper overexposure image on the left, you’ll find the base color RGB set to: R = 0.8, G = 0.05, B = 0.05
By slightly increasing your values to above zero, you’ll find that the lights in your scene will properly overexpose your objects.
As Turner says about the image with zero values,
On the right, we have some bad exposure. Here, no matter how much light we throw at the scene, no matter how much light we throw at that surface, it will never go white. It will never clamp, it will never overexpose in the correct way. It will always go to a pure red value because of the math underlying how it works.
He reiterates that whenever you’re making textures, you don’t want to use zero as an RGB value. In the same way you’d never want to use an all black 0,0,0 value that will never light up, keep in mind that same applies to RGB as well.
It’s a really solid introduction to the simple lighting math, and I highly recommend giving it a watch.
So if you’re finding weird results in your renders, check those values.
For a deeper dive into this, Turner suggests the following two papers.