Game Development Reference
In-Depth Information
3.5 Baked 3D Animations
Calculating animations in real time through kinematics is a very processor-
costly method for animating game characters. While this method can certainly
produce spontaneous interactions with characters, far more artificially
intelligent controls need to be programmed to make them feel real. If you
examine game characters closely you will notice that they repeat the same
actions in exactly the same way over and over. This is because it is less effort
on the part of the animation to make; for example, why use five different walk
animations when just one will suffice? It does not add anything more to a game
having characters that can select from different walking styles as they see fit.
If you stop to observe ancillary characters in the crowd such as those in
Splinter Cell or Grand Theft Auto , you will be able to spot their walk cycles
and other repeated actions. Because the purpose of the game in these
cases is not to eye the crowd, the same repeated walk cycle is not that
important. You may also find a game hero climbs a drainpipe the same
way he climbs a rope. Reusing animations is just a trick to get better
performance out of a game and a way to keep the development budget
down. In the end it is how these animations are used and how the game
environment is designed around these limitations. For example, the
animation for climbing a ladder could be used for scaling a building or
a trellis if the objects are designed to have handholds and footholds in
similar positions to a ladder.
When animations are fixed and not manipulated in real time by the program,
they are called baked . This means that the entire animation sequence
is calculated beforehand and that the program receives a set of model
transformations for each frame.
Because the character modeler cannot possibly know at the time of
animating how the player is going to direct and move the character, it is
impossible to create a long strung out animation, for example, showing
the character running up the road, jumping over a fence, and rolling
through a window. If the player decides the character should not jump
over the fence but rather hop on the nearest motorbike, a single long
animation will not allow for visualization. Instead, animations are broken
into short action segments that can later be put together in any order to
facilitate fluid animation. Such segments might include walk, run, and
jump cycles. Depending on how the player wants the character to
move, he can then run, jump, run, and walk or walk, jump, jump, run,
and walk seamlessly.
This means that each animation segment should start and end with the
character in the same pose. When the next segment is added, there is no
obvious gap in the sequence. As shown in Figure 3.7 , the initial frame for the
character's idle, run, walk, and shoot down animations has the character's
feet and legs in exactly the same position. No matter what the sequence of
Search Nedrilad ::

Custom Search