Sep 21, 2007

The Wonderful World of Multi-Pass

I'm not sure who is reading this blog. Every so often I find out that someone else is leafing through these entries. And while this blog originated as a place for me to show my family what the heck I'm doing out here (having fielded one too many, "What is it that you are studying again?" questions), to you non-family newcomer I say: welcome.


So what is multi-pass rendering? It is boring if you're not into that type of thing. And I don't want to bore you, I want to spark within you a new interest in the technical hurdles that must be leaped in order to furnish your sparkling little world of multi-media. I'm certain that the average joe gives no thought to how his media is created. And that's where I come in.

If you have seen Toy Story, Shrek, Cars, Madagascar, Polar Express, then you have seen 3D animation. These movies are made by literally modeling every nuance of the scene in a 3D environment within the computer. They actually take longer to make than most live-action and cell-animated (Snow White, Little Mermaid) features because modeling takes time. What's more, after it is modeled, it must be staged, animated, lit, textured, and finally rendered; the latter of which can take months to finish.

This process also applies to adding a 3D object or character into a live-action scene. The earliest and still one of the most prominent examples of this is Jurassic Park. Many of those Dinosaurs were CG (computer generated) elements and were not on the call sheet during production.

Now, back to the topic. The blessing (and curse) of 3D is that you have total control of the environment. This includes lighting, atmosphere (fog, haze), timing, blocking-- all things that live-action shooting may leave to chance or mother nature; and traditional cell animation would leave to: once it's drawn, it's drawn. But in 3D, you can build it, arrange it, rearrange it, light it, tweak it, color correct it, change the depth of field, saturate it, de-saturate it, darken it, lighten it, soften it, sharpen... you get the picture. There is absolute control over the final image.

Now I am a beginner, but let me show you an example:
This is a model I made two terms ago. I modeled it, textured it, lit it, and pressed render. This is the result of all the elements being shoved through the pipeline and
into a single frame called the "Beauty Pass". Because rendering takes so long, it is very tedious and time consuming to render, tweak, render tweak, etc.

So to avoid that waiting game we render each different aspect of the image separately. This gives us different images (called "passes") that are later composited together outside of the 3D realm. Remember, the final output is a single 2D image. Even if it is a moving image, the result is a 2D picture. And compositing a handful of 2D images is a much more controllable and quicker process than trying to render one perfectly constructed frame out of your 3D pipeline.

Here are a few different passes:

This is a diffusion pass:

This is pure color. No shadows, no highlights, no haze, just the color of the object.










This is a specular pass:
Here you get the highlights (the glint if you will) from the shinier metallic elements. These will be laid over the diffusion pass.









This is a shadow pass:
With this as a map, I can control the brightness of the shadowed areas. Because I have dark isolated from light, I can shift the brightness of the overall scene, but only affect the dark parts mapped out by the shadow pass. Make sense? Essentially, this pass gives me more control.







This pass is called the "ambient occlusion" pass:
This is one of my favorites. This pass helps define the overall depth of the scene. Lately there have been lots of commercials that feature animations with only this pass.









When you composite them all together, you have an image that is much closer to your goal. This is the final comp of the multi-pass render (the original beauty pass is below):

No comments: