Building Intelligence Around Video.
Video Spatial Intelligence
Exploring how AI can represent position, depth, scale, orientation, and spatial relationships between people, objects, and environments.
Temporal Consistency
Developing approaches to reason across time — tracking entities, actions, scene state, and continuity across frames and shots.
Physical Interaction
Investigating contact, support, collision, motion, occlusion, and object permanence to improve the physical plausibility of generated scenes.
Entity Consistency
Exploring methods for maintaining the identity, appearance, attributes, and relationships of people, characters, products, and objects.
Semantic & Instruction Adherence
Developing ways to evaluate whether generated video remains faithful to its intended script, actions, narrative, and creative direction.
Efficient AI Systems
Investigating specialization, optimization, and distillation to make video intelligence more efficient, deployable, and practical at scale.