mduffor

I made a posting over at CGTalk recently, and it seems to be useful to a lot of people.  I’m reproducing the suggestions here for your viewing pleasure. 🙂

1) Motivate your lights – As much as possible, every light in your scene should have a logical source to it. Is your key light coming from a lamp? a window? the sun? Sure you’ll wind up cheating the position and intensity of lights a bit for a better look, but try to keep them logically consistent

2) Don’t be afraid of the dark – Too much CG is over-lit (high key lighting). Let parts of the scene be under-lit, or drop into near black. You still want some shape in your dark areas/shadows, but it can be just the suggestion of forms. If you know an area will always be in near darkness, this is also an opportunity to save some time an not model/texture everything in this area to a detailed level.

3) Dark scenes do not mean under-exposed scenes – If you have a night-time scene or a dark interior scene, don’t just turn down the intensity of all the lights and call it a day. Drop the intensity of your fill, and even cool your fills off with a blue tint, but keep some elements of your scene properly lit. A character that is lit from the side (or slightly from behind) with a full intensity key and then darker fills will look better than a character that is more front lit but with the key turned way down in intensity

4) Depth of Field – We are used to looking at images taken by a camera, and camera lenses are unable to keep everything in focus at once. Items that are too near or too far from the focal point have softer focus. This can be done in 3D (slow to calculate) or 2D (faster, but with more artifacts that have to be worked around). Also remember that you need to use a circle/bokeh blur, not a gaussian blur if you are doing this in comp.

5) Keep your light sets separate – In real life, you often have to light your actors/subjects with at least some of the same lights that light your environment. But as a result you wind up having to set up all sorts of flags, diffusers, barn doors, scrims, etc. in order to keep the light at one intensity for one object in frame, and another intensity for other objects in the frame. If you separate out your environment and your subject where possible and light them with different sets of light, then you have more control over the lighting of each element. Each of your characters can be lit with their own set of lights as well. Most CG films I’ve worked on have a set of lights (key, fill, a couple of rim lights) for each character, and these lights are constrained to the center of gravity/placer of the character and move with that character (but don’t rotate.. they stay aligned to world space). This gives you a lot of control over each character’s lighting, and you can adjust one character without affecting the others. Also by rendering characters separately, you save a lot of time when you are tweaking lights. Then the characters are put together in the compositing program to create the final image.

6) Learn to use a compositor – Whether you are using Photoshop/Gimp for still images or Nuke/Shake/After Effects for image sequences, you should learn to use a compositing program. It is much faster to make color corrections, blur edges, fix render artifacts, adjust fog/atmosphere/fx-levels, adjust reflection amounts, etc. in 2D rather than tweaking settings in your 3D package and re-rendering. The faster you can turn around and see the results of a change/tweak, the more interations you can perform in a given amount of time. The more iterations you can pull off, the better you can make your art. This also ties into the above suggestion that you render layers separately so that a tweak that affects one element doesn’t require re-rendering them all. Don’t get caught up in trying to get perfect renders out of your 3D package. Get useable layers out of your 3D package, your 2D package, your particle engine, etc, and then blend them together in your compositor.

7) Linear Workflow – As others have mentioned, linear workflow is very important when you want lighting that you can control. Others have covered the how and the why of this elsewhere, so I’ll only mention it here.

8) Composition – Likewise, I’ll echo the sentiments of others and stress that image composition is very important. Make the meaning of your image clear. Light and compose to focus on your subject matter. Remove distractions from elsewhere in your frame. Especially if you have moving images, and your audience can only focus on the image for the few seconds (or fraction of a second) that it is on-screen, you want to focus attention and remove distractions so that your intended message comes across.

9) Calibrate your Monitor – If you have access to colorometers that can be used to adjust the color levels, brightness, and contrast of your monitor, then do it. Otherwise there are test images and procedures out there (Google is your friend), that will help you at least get your monitor somewhat close to calibrated. I’ve run across several renders that were too dark, and the reason turned out to be that the artist had the brightness on their monitor cranked soo high that they were able to see the image, but it was way too dark for everyone else.

10) Atmosphere – One of the things that make CG images look very CG is that their color stays the same across vast distances. In real life (assuming you are on a decent-sized planet’s surface) you view distant objects through a lot of air. This air has color and opacity that blocks/scatters the light moving through it. So if you are looking at objects more than a few meters away, you should add some sort of atmosphere to your render (usually best done in your compositing package). Render out a depth-from-camera pass (z-depth), bring it into your comp package, adjust the values with expand/gamma/curves, multiply it by your fog color, and add it back in on top of your image. This additional atmosphere really helps to sell that your image exists in the world, and gives it a sense of depth.

There, hope that helps in a constructive way. 🙂

Cheers,
Michael

Tags: CG

This entry was posted by mduffor on Friday, January 29th, 2010 at 4:45 pm and is filed under Computer Graphics. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.