July 9th, 2026 Product
List Cloud Egocentric API
See what the wearer sees
The Egocentric API turns first-person video from glasses, helmets, and bodycams into structured understanding: hands, handled objects, actions, and subtasks, timestamped and queryable.
Training-data annotation that took weeks now runs overnight.
Hand and object tracking
Fine-grained hand boxes and grasp states anchor every interaction — the foundation for imitation learning datasets and assistive applications.
Subtask segmentation
Long recordings split automatically into steps (pick, inspect, place), each searchable by natural language description.
Built for scale
Batch hours of footage through one endpoint; results stream back as JSONL ready for your training pipeline.