Pegasus 1.6 by TwelveLabs

Pegasus 1.6 is a video understanding model by TwelveLabs designed for robotics and physical AI teams. It analyzes egocentric and teleoperated footage from a first-person point of view, recognizing entities and producing persistent metadata across segment types. Teams can feed raw teleoperation or wearable video into the model and receive labeled, timestamped data ready for robot training, without a manual labeling pass.

Startups building in robotics and physical AI often sit on large volumes of raw first-person video that is expensive to label by hand. Pegasus 1.6 addresses this bottleneck by turning that footage directly into structured, timestamped training data, letting small teams move from data collection to model training faster.