Key Notes
- Pegasus 1.6 adds dedicated first-person video support and native image analysis.
- The platform prepares labeled footage for robotics pipelines rather than controlling robots.
- Video input is listed at $1.75 per hour, with output and Segment request charges affecting the total.
TwelveLabs has released Pegasus 1.6, a video-understanding model that adds dedicated support for first-person recordings used in robotics training and labeling pipelines. The update targets footage captured by body-worn cameras and robots’ onboard cameras, helping teams turn recorded activity into structured descriptions.
The company’s October 6 announcement also introduces native still-image analysis through the same API and improved recognition of people and objects. TwelveLabs says the model works with existing Pegasus SDKs, prompts and API integrations.
First-Person Video Becomes Training Material
Egocentric footage records a task from the perspective of the person or machine performing it. For robotics developers, that can expose useful details about hands, tools, objects and the sequence of actions, provided those details are labeled accurately.
TwelveLabs’ physical AI platform describes timestamped action segmentation, dense captions and configurable action taxonomies. A developer can define the categories a dataset needs, rather than accepting a single fixed labeling scheme.
The company also lists checks for camera stability, framing, occlusion and action clarity, alongside search for unusual events and duplicates. Its sensitive-content flags cover material such as faces, documents and badges. These are tools for screening footage before review and downstream processing.
An Update to an Existing Data Pipeline
Structured video descriptions are not new to the Pegasus family. In its September 30 technical overview, TwelveLabs described Pegasus 1.5’s temporal segmentation and metadata as a foundation for robotics annotations. The newer model’s distinct addition is support for first-person footage, a category the company says earlier versions did not handle well.
CEO Jae Lee told VentureBeat that human video could complement teleoperated robot demonstrations.
That places Pegasus upstream of robot execution: it describes and organizes observations. Developers still need systems that connect those observations to robot movement, sensing and control. TwelveLabs’ own overview places its descriptions ahead of the geometric work needed to recover poses and contact points.
Pricing Depends on the Annotation Request
TwelveLabs’ published API rates list Pegasus 1.6 video input at $1.75 per hour, image input at $3 per million tokens and output at $7.50 per million tokens. Enterprise pricing is custom.
At the listed video-input rate, processing 1,000 hours once would cost $1,750 before output and other applicable charges. Segment requests have an additional billing rule: each segment definition is charged separately, multiplying the duration billed. A request covering several action categories can therefore cost more than a single pass over the footage suggests.
The Test Is Better Robot Training
The data-preparation approach complements other work covered by RobotsBeat, including Penn’s SymSkill research on learning reusable movements from demonstrations and Bonsai World, which creates simulated environments for field machines.
For teams evaluating Pegasus, useful measures include labeling accuracy, missed action boundaries and human correction time, followed by whether accepted data improves the target robot task. Producing more annotations is a different result from demonstrating more reliable physical behavior.
Disclaimer: RobotsBeat is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, Robotics, technology, software, and digital innovation sectors. These relationships do not influence RobotsBeat's editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
