Synchronized Multimodal Physical AI Dataset - POV Video, Eye Tracking, Full-Body and Hand Motion Capture, and Force Sensing | World Model Data

This world model dataset is produced on demand based on the buyer’s required tasks, sensor stack, participant profiles, capture environments, annotation depth and target volume. A technical pilot can be created before full-scale production to validate data quality, synchronization accuracy and sensor performance. The dataset combines synchronized first-person RGB video, eye tracking, full-body motion capture, detailed hand and finger tracking, grip-force or pressure measurements, audio and structured task metadata. Participants perform physically meaningful manipulation tasks using real objects, tools and environments. Task categories can include household activities, assembly, warehouse operations, retail workflows, object sorting, picking and packing, equipment handling, maintenance, tool use and fine motor manipulation. The dataset is designed for Physical AI, humanoid robotics, embodied AI, vision-language-action models, world model data generation, robot imitation learning, dexterous manipulation, behavior cloning, grasp prediction, force prediction, gaze-guided action understanding and human-to-robot motion transfer. Each recording session can include: - First-person RGB video - Gaze coordinates, fixations and saccades - Full-body skeletal motion - Hand and finger motion - Grip-force or pressure measurements - Synchronized audio - Calibration data - Sensor health data - Session and task metadata - Cross-sensor synchronization metadata Optional synchronized external came