Lander, Jeff (Darwin 3D LLC)

Thema: "Developing High Fidelity Animation Technology Without Going Insane"

Jeff Lander is founder of Darwin 3D, a company geared toward a higher adaptation of real-time 3D Graphics. Jeff has worked as a programmer for over 20 years in the video game, television, and film arenas.

He has worked on many interactive 3D projects for a variety of clients including nVidia, Rhythm & Hues Studios, Sony, Ion Storm, Exakt Entertainment, Electronic Arts, and Activision. He also wrote the monthly Graphic Content column for Game Developer Magazine from 1998-2001, co-founded the Game Technology Seminars, and has edited several game programming books. He is a frequent speaker at game conferences and also teaches an advanced game programming course at the Art Institute. He is currently working on an unannounced game console project.

Summary of presentation:

Game animation systems are largely still stuck in the mode of playing back and blending between linear clips of animation data.  There has been movement to improve the low level blending techniques and production pipelines.  However, if we are to scale our systems to handle the level of fidelity needed to create more realistic worlds, we will have to start changing our approach.  Animation state machines and blend trees are great for getting something up and running quickly.  It becomes quickly apparent that these systems do not scale well once you begin to layer in the complexity of emotional and motivational performance that games need to start creating.  Animation middleware companies have begun to make tools to make these classic animation structures easier to edit and maintain.  But this session will argue that these methods are attacking the wrong problem entirely.  Additionally, the research community has provided a lot of ideas for low level improvements in our systems.  However, these are not tested and proven in actual production work.  In this session we will lay out the design for complex and scalable animation systems that combines a progressive-refinement approach to low level complexity and fidelity with an animation database system that scales well without the brittleness of a blend tree.  Using the kind of difficult production scenarios that game engine companies never talk about, the session will cover approaches to solve those dilemmas.  Attendees will be presented with the current state of the art for game animation systems and provided with a roadmap for way to integrate the new techniques discussed into their projects immediately.  The session will cover animation technology at a very deep level with some programming examples as well as art production tools.  Attendees should be highly familiar with traditional animation techniques such as animation blending, anim trees, and kinematic and dynamic animation.

Attendees will gain an overview of the current state of the art in animation systems in research, production, and middleware.  The strengths and problems with each of these will be discussed.  A new approach to creating high fidelity animation that is scalable for large projects will be presented.  These ideas will be demonstrated in game production code and can immediately be applied to the attendees own projects.  The session is targeted toward animation programmers, technical directors, technical artists, and animators who design the animation systems and pipelines for their projects as well as use them in production.