China puts AI video on an assembly line, a third cheaper per minute

China stopped treating AI video as a picture generator. On November 4, Orca Entertainment launched Jingrui AI — a platform that turns AI video into an industrial line. The bet is not on generating something pretty, but on shipping a project on time and on budget. That is where the real boundary sits: competition moves from the quality of a single frame to control over the whole process.

What's inside the pipeline

• More than fifty third-party video generation models and about ten in-house ones under one interface.
• An infinite canvas: thousands of nodes, a hundred people working on a scene at the same time.
• A finished minute of AI video costs over a third less than with ordinary generation.
• Sign-up without a subscription, credits bought on the fly, no cap on team seats.

The seat cap matters more than it looks. Studios hate paying for someone who joins a project for a week. Kill the seat counter and you kill the main excuse for not bringing in outside artists.

The problem it actually fixes

Chinese creators call it gacha: generation is a lottery. You keep spinning until a frame lands, burn credits, lose a day. A pipeline exists to replace the lottery with a brief. Hence the scene-as-graph model: every node is a specific shot, a specific model, a specific fix. Change one node and you don't rebuild the entire film.

Regulators counted 430,000 micro-dramas launched in the first eight months of the year, and more than nine in ten of them were AI-made. Footage is no longer scarce. Delivery is. The scarce thing is a whole project that lands cleanly: the scene doesn't regenerate from zero, a fix doesn't break the neighbouring shot, the budget doesn't vanish into one more take.

Jingrui AI aims straight at that. Orca CTO Du Yanlong says the platform will be polished on real projects and that teams should bring harsher demands. Translated from corporate: features aren't enough, orders are. Volume is what proves the promise.

Where it's thin

The platform has no generative magic of its own — it aggregates other people's models. It's a well-managed project layer, not a new foundation. The moment any lab ships a model that holds scene and character consistency by itself, the wrapper drops to being a nice interface.

And the word first is written by the owner. Nobody has independently benchmarked it against Western AI production pipelines yet.