Artistic Problems

Creating convincing and appealing facial expressions, like smiling or blinking, seems simple in concept. But, it requires deep understanding of the incredibly complex system that lies beneath a face and very good knowledge of animation principals, to perfectly reproduce facial movements that look realistic. Such realism is constrained by two main factors:

• inconsistency of facial movements: humans are very talented in identifying unnatural behavior, due to their everyday familiarity and sensitivity to facial appearance. Non-realistic animations will generate in the audience unpleasant and sometimes repulsive feelings, disturbing the visual perception;
• diversity of faces: human faces have different facial features that emphasize their uniqueness, caused by the variation in size and proportions of their bones and muscles. On the other hand, cartoons or fantastic creatures might not follow a human structure definition, which increases the diversity range.

The uniqueness of faces makes facial synthesis so challenging. The smallest anomaly in the face shape, proportion, skin texture or movement is immediately detected and classified as incorrect (this is known as the Uncanny Valley hypothesis).

Technical Problems

The rig allows artists to edit the model using a set of controls, instead of having to position each vertex of the surface on the desired location. Today, most rigging systems present some of the following technical problems:

• no standard: artists do not follow a formal criteria or methodology when creating a rig; they do whatever “feels right”, so all rigs end up being different. Defining a standard would help create a solid platform for artists to build upon;
• changing the geometry or resolution: it is very common to change the face model during production, to improve the deformation details or simply because it looks better. Any minor modification in the model surface (a bigger nose, more resolution around the lips) after the character is rigged, causes the rigging process to restart;
• reusing weight maps: the weight distribution defined for one character will not work on others. If it is not possible to reassign the appropriate weights in the new model, the resulting animation will present undesirable deformations, especially if the models differ in geometry and proportions;
• number of shapes: many productions use rigs based on hundreds of shapes. Usually, too many shapes make it hard to use the rig. Likewise, if a shape is added during production it can generate two problems: the shape conflicts with existing animations, making it necessary to rework some shots; or the new shape does not mix nicely with the the others;
• preserving a consistent look: placing by hand the animation controls leads to different artistic interpretations of where to position each element of the rig. This makes it difficult to easily reproduce the same facial pose between different characters. Consequently, it becomes hard to guarantee a consistent look throughout the production;
• complexity of the UI: artists place many controls to handle every region of the face and reproduce the subtleties of each expression. If controls are not efficiently organized in a few layers, it can lead to an extremely complex UI with a steep learning curve. It will take animators too much time to master the UI;
• adding rig controls: a rig usually goes through multiple iterations before an animator starts working on it. It is very common to change animation controls or add new ones during production. This leads the rig to be readjusted and tested, to verify if it still deforms correctly.