Creativity tools for digital media have been largely democratised, offering a range from beginner to expert tools. Yet computer animation, the art of instilling life into believable characters and fantastic worlds, is still a highly sophisticated process restricted to the spheres of expert users. This is largely due to the methods employed: in keyframe animation dynamics are indirectly specified over abstract descriptions, while performance animation suffers from inflexibility due to a high technological overhead. The reverse trend in human-computer interaction to make interfaces more direct, intuitive, and natural to use has so far hardly touched the animation world: decades of interaction research have scarcely been linked to research and development of animation techniques.
The hypothesis of this work is that an interaction approach to computer an- imation can inform the design and development of novel animation techniques. Three goals are formulated to illustrate the validity of this thesis. Computer animation methods and interfaces must be embedded in an interaction context. The insights this brings for designing next generation animation tools must be examined and formalised. The practical consequences for the development of motion creation and editing tools must be demonstrated with prototypes that are more direct, efficient, easy-to-learn and flexible to use.
The foundation of the procedure is a conceptual framework in the form of a comprehensive discussion of the state of the art, a design space of interfaces for time-based visual media, and a taxonomy for mappings between user and medium space-time. Based on this, an interaction-centred analysis of computer animation culminates in the concept of direct animation interfaces and guidelines for their design. These guidelines are tested in two point designs for direct input devices. The design, implementation and test of a surface-based performance animation tool takes a system approach, addressing interaction design issues as well as challenges in extending current software architectures to support novel forms of animation control. The second, a performance timing technique, shows how concepts from video browsing can be applied to motion editing for more direct and efficient animation timing.