Kiel Figgins demonstrates and explains the key techniques in creating dynamic, fast-action character animation
This video shows the boards provided by the client and their initial 3D blocking (the sepia tone). You can see how a T-Pose character doesn't convey the scene very well, plus with the uninformative environment and viewport clutter, this meant their blocking had to be redone from scratch. You can see where I picked up the shot around the 10 second mark. As the ideas started to take shape, the shot grew longer and more story points were added. The following tutorial is based around that process.
Showing the stages of a fast action shot
One of my favorite types of animation is fast action and acrobatics. These shots have so much energy and excitement that keying the flow and poses really lets you push what the character's capable of.
When I'm assigned a shot like this, my typical approach is to make a proxy of the character and environment, and animate the entire sequence on those proxies first to get the flow. Since these shots are more about the transfer of inertia, I don't worry about the poses until later. These shots typically require a lot of iterations, revisions, or complete overhauls to get them to work in the edit, so speed and playback speed is a must. Working with proxies allows for timings to be made in the viewport with minimum playblasting, while keeping the keys clean enough to slide around the major actions.
On another level, having the core motion on a basic, non-descript, character allows the viewer to project their own mental image of the character onto the action. The motion doesn't look "wrong", since it's not a real character. This approach stems from getting an idea across in instances where stepped poses wouldn't clearly be understood or applicable (Previz, motion graphics, game engines) or when working with a client that is unfamiliar with stepped poses and may view them as a compression error, glitch or mistake.
Cube jumping through buildings
Cube jumping into a moving car
A common practice is cutting up the high res mesh of the character and using those pieces. With that approach the viewer will keep looking at how the mesh is chopped up, and worst of all, the static expression on their face. Having basic cubes allow you and the viewer to focus solely on the motion, without the distractions of penetration, bind issues and expressionless faces.
There are several benefits to this approach:
Separation between timing and making beautiful poses
. Focusing on one element at a time leads to a more concise final product. If I have to worry about the poses at the same time as the flow, I typically get bogged down. Also, due to the highly iterative process of fast action shots, creating rapid versions that can be seen together in context of one another (or in the edit) is desirable. If the flow, timing, or staging needs to be changed, having the shots in a malleable state reduces time lost crafting poses that may not be used or scrapped entirely.
Able to hand over the blocking scene to other Animators without clean up
. Every Animator will set a pose differently with the same rig. Since the blocking has been done on a proxy element, the Animator, either in house or remote, can set the pose as they see fit. The Animator will have the proxy animation to refer back to, also, if they've referenced in the blocking scene, the blocking can be updated later as needed. There becomes no need to delete keys, work off existing keys, or have multiple references of the same character. Often times, the final rig isn't even completed at this early stage, so this allows animation to start while the rig is in development.
Final assets, such as models or rigs, are not needed for animation to begin
. Often a Director will have an idea of a sequence or moment that he wants to see quickly. It may be a new addition or a change in an existing shot. Being able to quickly block out the key elements not only shows that an idea will or won't work, but can provide a base for scale for modelers, and range of motion of the character for the Setup Artists to have a heads up on.
Less personal connection to the poses
. The mentality of adjusting 10 controls to change the pose is far more forgiving, compared to all the controls of a full rig. Plus, if animation needs to be duplicated, offset or transferred to another character, there are far fewer controls to worry about.
Able to determine if distance travels is applicable to the character
. Often Animators will "skate" a character in T-pose around a scene to illustrate him walking or running. Unfortunately, this does not tell the viewer if the legs are going to be moving too fast. If they are, the distance traveled will need to be adjusted, either in the type of locomotion or by changing the environment. Both of those can drastically affect the mood of the shot, imagine a femme fatal striding through a doorway, now imagine her having to jog to cover that distance.
Animation blocking is more descriptive if someone else is doing the camera or building the environment around the animation
. Having the arcs and full frames of motion is far more useful if someone else is doing the camera work for your shot. Knowing where the characters are in space affect the composition more, than the intricacies of the pose. Plus, if a character is running through a city, the environment artist will have some idea how much of the street to flush out and what elements the character or camera gets close to.
Stepped poses in game engines read as dropped frames
. A stepped animation in a game engine can read as many things, such as a playback issue, export error, frame rate drop, or a heavy compression setting. They also are not as useful to designer that may need to work with blending the animations or working with it for scripted events.
Setting up the character
Building a proxy of your character should be a fairly short process. It's all about approximation, so leave out: hands, fingers, feet, toes, face, and accessories. If your character has a defining feature, such as a large jaw, and is biting in the animation, proxy that in as well. The goal here is get a character that is quick to animate and playback in real time in your scene
Number of controls per layer
This character would have no controls, only parented polygons, all channels locked if they cannot be keyed (typically scale and visibility), and all placed under a single group. The proxy is typically made of separate poly cubes, roughly sized and shaped to the limbs of the character, with pivots at the logical bend points (knees, elbows, etc ). If you can match the pivot position of a control for the proxy mesh, that's a bonus. If you plan on transferring keys, be sure to match the rotation order of the control. Don't spend too much time here, since you'll be redoing those poses on the final rig anyways, so it's best not to noodle. To save on modeling more cubes, I typically allow translates on the upper arms, legs and head to act as clavicles, hip and head tilts.
Hierarchy of proxy character