Hollywood in 18 Months: From Banning GenAI to Demanding an AI Strategy from VFX Studios

As recently as early 2025, the phrase "no GenAI in our projects" sounded like the norm in Hollywood contracts. By autumn 2026, it is the reverse: if a VFX contractor has no coherent AI strategy, a studio may not even start discussing a project with them. Variety assembled this evolution from the words of Weta FX head Daniel Sia, and the Telegram channel "Metaversishche" retold it with some irony.

Four Stages of Acceptance

Sia describes the path as follows (paraphrased, not verbatim quotes). Early 2025: "no GenAI." Mid-2025: "allowed, but only with permission." Early 2026: "how much will we save on this?" Autumn 2026: "no AI strategy, no conversation."

The trajectory is a familiar one. First legal fear, then cautious permission, then the budget question, then a standard. What matters here for a product manager or founder is not the irony but the shift in logic: AI has stopped being a risky option and has become a criterion for selecting vendors.

What Weta Is Building and Why It Isn't Just Another Generator

The most interesting part is how the studio approaches the subject. Weta FX, which did the effects for "Avatar" and "The Lord of the Rings," is developing its own specialized models for modifying and creating digital worlds. These are tools for producing visual effects, not a "make a clip from a prompt" button.

Together with Taiwan's Moonshine Studio, Weta is building tooling that lets artists control generation across an entire film. The problem is clear to anyone who has tried to assemble a long video from generated clips. The model produces 20-second fragments, while in a two-hour film the characters, style and logic of the world must hold together. Stitching such pieces together by hand is not a pipeline but a lottery.

In parallel, the open-source project RMTC is being developed (it lives in the Academy Software Foundation organization on GitHub). It is meant for managing model training, datasets and content rights. The idea is pragmatic: in film, the question "what was the model trained on and are the rights cleared" matters more than how pretty the picture is. Without traceable data, a studio cannot explain to a client or to lawyers where a result came from.

Kling AI and Seedance are already used in real production, but only where clients allow it. In addition, Weta plans to buy a substantial stake in Moonshine Studio and is taking part in creating an AI hub in Taiwan. In other words, this is not an experiment in a corner but an investment in infrastructure.

Peter Jackson's Position

Back in May in Cannes, the director of "The Lord of the Rings" said he has nothing against AI in film: to him it is just a special effect, no different from others. The only firm line, in his words: digital copies of actors must not be used without permission.

Weta FX management, for its part, insists that the AI pipeline is being built to expand artists' capabilities, not to replace them. This is the studio's own statement, and it cannot be verified from a single publication.

What This Means in Practice

If you are a service provider in the creative industry, an AI strategy becomes part of your commercial offer. It needs to answer several questions: which tools are used, where the client has permitted this, how data rights are handled, and who makes the decision on each project.

If you build tools, demand is shifting from "yet another prompt window" to control over long-form content. What is valued is consistency across an entire film, dataset management and rights tracking. The last of these is exactly what RMTC covers, and that piece could well become a standalone product.

If you are a product manager, pay attention to the formula "when clients allow it." Client-level and project-level switches for AI permissibility, along with a log of what was generated and with which tool, are turning from a nice-to-have into a mandatory part of the workflow.

Limitations

The source is a retelling of a Variety article in a Telegram channel, with the author's jokes and without figures. We do not know how exactly studios phrase the "AI strategy" requirement or what counts as a sufficient answer. There is no data on savings or on the share of AI-generated shots in finished films. The plans to buy a stake in Moonshine remain plans for now. Still, the direction is clear: in an industry that recently banned generative AI outright, the question is no longer "whether to use it" but "how, with what controls and with what rights."