Character Animation

How can digital avatars move in ways that are realistic, adaptable, and purposeful to meet users’ demands? Character Animation research aims to develop learning-based methodologies that go beyond conventional animation techniques such as keyframing or motion capture. Our approach enables avatars to respond dynamically within interactive environments while remaining easily controllable by users. This research has wide applications in gaming, immersive virtual worlds, and digital media production, where interactive and expressive character motion is essential.

We explore methods to tackle motion generation, editing, retargeting, control, and planning. Our systems empower avatars to exhibit natural movements in dynamic environments, express diverse stylistic nuances, and execute user-driven actions seamlessly. We build frameworks that integrate large-scale motion data, generative models, and physics-based simulation into cohesive learning-based methodologies. This foundational work paves the way for developing immersive and intelligent experiences with virtual avatars.