Decart's Lucy transforms live video in real time — which is a claim about chips, not pixels
Lucy processes live video feeds to produce real-time footage — showing people wearing clothing or accessories they are not wearing. The application is fashion e-commerce. The engineering claim underneath is that the model hits frame rate on hardware that normally cannot.
Most generative video is asynchronous: submit a prompt, wait, receive a clip. Real-time transformation of a live feed removes the wait, and the wait is where every efficiency shortcut normally hides. Frame rate is an unforgiving benchmark because it cannot be met on average.
Virtual try-on is a legitimately hard target rather than a demo. It requires garment geometry to track a moving body, lighting to stay consistent, and occlusion to resolve correctly — frame after frame, with no opportunity to re-render a bad one. It is also a problem fashion e-commerce has been paying to solve for a decade.
The strategic reading is about the chip work rather than the clothes. Decart pairs its models with software that lowers training cost by improving chip utilisation, and real-time video is the demonstration that the utilisation work is real. You can claim efficiency in a benchmark; you cannot fake 30 frames a second.
Which explains where the team would land. In Anthropic's inference and performance organisation — not in a video product group.
TechRepublic — Anthropic Reportedly Eyes $6 Billion Decart Acquisition to Boost AI Efficiency → · TradingKey — Anthropic to Buy Decart AI for $6 Billion to Boost Video Generation and Chip Optimization Capabilities →