Nua

Nua is a video decision model product backed by Y Combinator. Users ask questions about their video footage, and Nua finds where a specified moment occurs and makes a judgment about whether it happened, for how long, and whether it passed. The page shows examples in robotics training data labeling, insurance claims footage, and sponsorship verification. Artificer, described as a 4B robotics model, is presented with benchmark results on ArmNet268 (92.84% balanced accuracy) and RoboReward (0.4308 source-averaged mean absolute error). The page invites visitors to request access and submit a sample video with a question, and links to API docs.

Key features

• Finds specific moments in hours of video and pinpoints when they occur
• Makes yes/no judgments about events in footage, including duration and pass/fail outcomes
• Artificer 4B robotics model for judging task success and progress from video
• Benchmark results reported for Artificer on ArmNet268 and RoboReward
• API documentation available

Best suited for

• Robot evaluation
• Training data labelling for robotics
• VLA training
• Insurance claims footage review
• Sponsorship and co-op placement verification
• Advertising video analysis
• Fleet, sports, manufacturing, construction, and retail footage review

Pricing

Not stated on the website