Generative and Creative AI

We conduct research on generative AI methods for image, video, and 3D creation.

Topics

Diffusion transformers, flow matching, and architectures for high-fidelity image and video generation.

Controllable, temporally coherent video generation and editing from text, audio, and visual prompts.

Reconstruction, generation, and animation of objects, avatars, and dynamic scenes.

Identity-preserving generation, attribute-specific control, and subject-driven editing.

Text-to-motion generation, human-scene interaction, and bimanual animation.

Instruction-guided editing, temporal consistency, and egocentric video understanding.