An issue which is often overlooked or avoided by individuals new to 3D is the necessity of fine-tuning projects in conjunction with a compositing package. Yet the level of interactive flexibility available within just about any 2D application can save precious time. While it is possible when working on full CG projects to render the final image directly from the 3D rendering engine, this methodology is rarely utilized in production environments. Furthermore, as most work in broadcast and film utilizes 3D imagery as a subservient element to a live-action backplate, compositing is going to occur regardless. While many visual effects shots incorporate 3D, all visual effects shots are composited. What also must be mentioned is that there are many effects which not only benefit from, but require a compositing application's specialized features: Color correction, film grain, effects such as depth of field, fog, glows, motion blur, heat distortion, optical effects... while some of these can be achieved in Maya with varying degrees of success, an experienced compositor with a strong application can often times take things further.

My purpose here is to give an overview of how 3D elements are prepped and rendered from Maya for use in a compositing pipeline... You will notice, however, that I do not mention the Render Layer Manager. Reason is, as you may expect, that it is clunky and not really the best way to approach passes/layers.

Passes

While it is possible for an image rendered with Maya to achieve the results you need, the most common pipeline involves getting things are close as possible in Maya while keeping in mind what types of tweaks can be more easily made in a compositing package. Any serious lighter should be familiar with at least one application. This includes Nothing Real Shake, Adobe After Effects, Discreet Combustion, Silicon Grail Rayz and others.

Diffuse/Beauty Pass - rendered by turning off 'emit specular' on all lights.

Specular/Highlight Pass - rendered by turning off 'emit diffuse' on all lights.

note: either Diffuse or Specular passes can even be rendered on a light by light basis to increase the degree of post-control. This is should not be necessary very often, although it is not uncommon in film.

Reflection Pass - Create a chrome shader (blinn, color=black, diffuse=0, specular = white, reflectivity = 1) and assign it to the object(s) being rendered. Reflection intensity/opacity/blurring can then be handled at the compositing stage. The objects being reflected will need to be matted out by rendering another matte pass.

Matte Pass - If an occlusion matte is needed, the occluding object's shader can be edited so that it's function in the alpha channel is to block out objects behind. This is done by setting the Opacity Control on a Shader to 'Black Hole'.

Shadow Pass - There are two types of shadows when rendering passes. The shadows an object casts onto itself (self-shadows) and the shadows the object casts onto other objects (cast-shadows). Usually just the shadows being cast on other objects are rendered as a pass, and the self-shadowing is included in the beauty pass. To render a shadow pass, turn off 'primary visibility' on the object's shape node, but leave 'casts shadows' on. One can also assign the 'Use Background' Shader to objects receiving the shadows so that only the shadow is rendered (to the alpha channel).

The issue of shadow accuracy on a 'hard - soft' scale can be dealt with in two ways. One is to use d-map shadows for self-shadowing which easily creates varying softness, but then render hard accurate shadows for cast-shadows so that complete control is taken at the compositing stage. It is easy to soften a hard shadow via gradient blurs, but you can not harden a soft shadow. The other technique is to try to get Maya to render accurate soft shadows but this can slow down rendering. If softening the shadows in post is going to be too arduous, then this will be the best route: it is always a question of which method is faster. At film houses it is very common to provide hard-edged but accurate shadows to the compositor so that all blurring/opacity/color correction can be handled in post.

Effects Pass - Elements such as fur, smoke, rain, etc are usually rendered as separate passes.

Depth Pass - Z-depth can be rendered as a separate pass to allow for a variety of results in post. One can use Z-depth to add depth-of-field and fog to an element. It can also be used to allow a compositing package to know where one object on a layer is in relation to another object on another layer. Therefore if you had a sphere in the middle of a torus where some of the sphere is behind and some in front of the torus, you could render the objects separately and still composite them successfully.

Z-depth is a single channel image, being limited to 256 shades of gray in a standard 8-bit/channel image. It is common to render a depth pass as a 16-bit image to increase the value range and accuracy within a z-depth image.

The gray values within a depth pass represent distance from the camera where white is near and black is far. The range in Maya units that this represents is based on the rendered camera's clipping planes. Therefore you can increase accuracy by using near/far clipping planes which are as close to the objects in your scene as possible while not clipping them off. It is very important, however, to use the same clipping plane values for each rendered layer.

Layers

While pass rendering allows you to improve the realism of an image, rendering objects in layers allows you to tweak objects independantly of each other. It also allows you to more efficiently manage render time. If a camera is locked down, for example, the background could be rendered as a still image with the animated elements composited on top.

To render objects separately, there are a couple techniques you can use. Managing the 'primary visibility' attribute via the Attribute Spread sheet is the most straight forward approach. You can also use either scene layers or Maya's render-layer manager, but I prefer to rely on scene layers. Just be careful when rendering layers via visibility to watch out for how those objects interact with each in regards to shadows and lighting.

Conclusion

So at this point it should be clear that once you combine rendering in layers with passes, a shot can easily be broken down into many elements. If separate light passes are also rendered, then the amount of elements can reach high numbers... so if you have heard case studies from films such as Final Fintasy or Titanic, where scenes sometimes have dozens to hundreds of composited elements per shot, hopefully this overview has given a clearer picture as to why there are so many.