At first, we were thrilled. As character animators, we couldn't have asked for a better project. There were two heroes, dozens of enemies, scores of NPCs, and more than 100 character-driven cutscenes. Enthusiasm and artistic latitude made it all ours for the taking.

But staying true to our shared vision of Ratchet & Clank meant that our digital actors needed to become more than mere cycling automatons. We regarded each character as an intermediary through which we could reach out to players and draw them deeper into our universe. This meant our characters needed to blend physically into their environments, emotionally into their situations, and expressively into our narrative. It was on these principles that we based both our objectives and our standard of success.

Our team acknowledged that a rift existed between the level of complexity we desired and the time we had scheduled to implement it. In order to surmount this obstacle, we developed several methods for using Maya, our artistic skills, and our time more effectively.

This article will discuss these methods both in terms of their functionality and their implementation. To this end, it will provide technical details on our testing practices, our MEL shortcuts, and our real-time animation procedures.

Furthermore, it will explain how each of these methods saved us valuable production time, enabling us to achieve our artistic goals.

Part of achieving our goal of tying our characters closely to their environments and gameplay meant prototyping low-resolution versions of our characters and their respective animations. Like coalmine canaries, we sent proto-models into our new levels to nose out potential animation, programming, and design problems. We relied on prototyping throughout the course of our production as a means of refining a character's move set. This process of refinement was key to winnowing down unworkable ideas before animating a character's high-resolution incarnation.

As a rule, our prototypes emphasized function over style. And although we set the aesthetic threshold low, these previsualization models still needed to be built and animated accurately enough to function as valid test cases. For the animators, this meant that prototype characters needed to jump to their correct heights, attack to their design specifications, and run at their proper speeds.

Generally, we created prototypes using a character design sketch as a guide. These proto-characters were constructed with primitive objects and only roughly resembled their future incarnations, as you can see in Figure 1. Since previsualization models were so simple to construct, every animator could assist in building them, regardless of their modeling experience. Accuracy was required only in the representation of the character's height, proportions, and posture.

	Figure 1. The Dog Charger prototype was used to pretest the final character's animations, including its walk, run, and attack.

For the most part, our prototypes had extremely simple skeletons: all geometric components were assigned to a single bone with no special deformation. Though such simplicity made for blocky-looking models, in practice our animators had all the flexibility they needed to test out a move set.

Animating our proto-characters was similar to sketching a traditional pencil test. Although animators were given a designer-approved move set, it was understood that animations needed only to be rendered into their roughest forms. One pass was often sufficient, as polish and overlap were unnecessary.

The areas where precision did count were timing, measurement, and interaction with other characters. As they have the greatest direct impact on gameplay, these attributes were considered critical to testing a new character's behavior accurately.

Timing has a major effect on both the readability of an animation and on gameplay. From a distance, a poorly timed idle can look muddy. An attack animation can be too slow to make an enemy a worthy opponent, or too fast to be registered. Emphasis or a lack thereof on just a few frames can make or break any animation, especially within the short cycles of the real-time universe we were creating. We discovered that by testing and fine-tuning our timings in the prototype stage, we could often avoid reworking polished animations on final characters.

Making sure that proto-characters adhered to design measurements was also important. For example, if the design document called for an enemy to attack at a range of 4 meters, animators would ensure that the prototype did exactly this. Designers could then get an accurate idea of whether an enemy traveled at the correct speed, was tuned to the appropriate difficulty, and was scaled appropriately in relation to the main characters.

Prototyping also gave us a means of pretesting character behaviors and interactions. Whether it was with Ratchet or Clank, with the environment, or with another character, proto-models provided invaluable early glances at interactive behavior. For artists, programmers, and designers, previsualization served to telegraph character behaviors both in terms of their technical feasibility and their gameplay value.

	Figure 2. The final Dog Charger model.

Ultimately we found that our previsualization process was beneficial not just to animators but to our design and programming staff as well. It gave our programmers a head start on coding gameplay, while designers could test, tune, and ask for changes at a very early stage, allowing room for refinements.

Prototyping saved animators time and energy that otherwise would have been spent painstakingly modifying or redoing final multi-pass animations. It provided a relatively simple means for evaluating character behaviors with respect to their timing, specifications, and interactivity. Moreover, it provided our animators with a practice run, complete with feedback, before moving on to a high-resolution character (Figure 2).

MEL Shortcuts: Automating Our Setups

Maya Embedded Language (MEL) scripts were essential for bridging the gap between the level of complexity we desired and the time we had scheduled to implement it. Through MEL scripts, we were able to streamline setup operations, customize animation processes, and level our technological playing field.

	Figure 3. This leg setup was used for most bipendal characters, saving tedious hand-setups for IK systems for individual characters.

Two such scripts (examined later in this article) allowed our team to take advantage of driven key functionality that otherwise would have been too cumbersome to animate or too tedious to rig by hand. Another tool enabled our artists, regardless of technical experience, to fit characters with IK systems automatically.

Most of our bipedal characters had leg setups like the one pictured in Figure 3. As seen in the hierarchy (Figure 4) our legs had standard hip, knee, and ankle joints, a heel joint, and two to three bones in the feet. (For clarity purposes, please note that we referred to our foot bones as "toes.")

Our IK-rig consisted of three to four RP (Rotate Plane) IK-handles. These connected hip-to-ankle, ankle-to-toe, toe-to-toe and/or toe-to-null. All were configured into a hierarchy (Figure 5) that specified relationships between the IK-handles, a set of locators, and several NURBS constraint objects.

	Figure 4. Standard hierarchy for a character's leg, as shown in the Hypergraph.

Though relatively simple, setting this IK-system up by hand for every NPC, enemy, and prototype would have taken more time than we had. Moreover, we knew that this time would be better spent bringing our characters to life.

An actual tools programmer might scoff at the artist-authored MEL script we developed to make our leg chains. In the end, however, our "IK Setup Tool" reduced an hourlong technical chore to a simple task that took seconds. Furthermore, the script did not require setup expertise, and our relatively simple code could be customized and refined entirely from within the art department.

Using the IK Setup Tool (Figure 6) was a three-step process. First, an artist checked their characters' leg joint names against the tool's presets, making any necessary changes. Next, a scale factor for the constraint objects was entered, based loosely on a character's size. The artist then hit nine buttons in sequence. These buttons would auto-attach the IK handles and instantly build the constraint hierarchy.