Hollywood and Generative AI: Why Weta FX Is Now Demanding an AI Strategy

TVS News AI, editorially supervised, presents Hollywood and Generative AI: Why Weta FX Is Now Demanding an AI Strategy.

On 5 November 2026 at 10:46 p.m. PT, Variety published a clear warning from Hollywood: Anyone seeking to work as a VFX service provider for major studios today needs an AI strategy. The statement did not come from a general study, but from Daniel Seah, CEO of Weta FX, the VFX company known for “The Lord of the Rings” and “Avatar.”

“If You Don't Have an AI Strategy, Don't Bother”

Seah's wording sounds blunt. Behind it, however, is a concrete change in the business relationships between film studios and their suppliers. According to the report, Weta's customers are primarily the big four studios and streaming platforms. Previously, they issued clear legal guidelines. Now they are no longer asking only about the use of generative AI, but about a robust concept for deploying it productively.

From “No AI” to “Show Us Your AI Plan”

The first prohibitive step is already behind the industry. Seah said that earlier in the previous year, studios informed him through their legal departments that there was no possibility of using generative AI for their work. By the middle of the year, several customers relaxed their position: They allowed its use, provided Weta obtained prior approval and worked only with a small number of selected companies.

Around the turn of the year, customers then began asking Weta directly about its AI strategy. Seah identified the main area of interest as how much money the technology could save. Since the third quarter of the current year, his message had become considerably harder: Without an AI strategy, working with these customers would not be an option.

The standard for review has therefore shifted. A ban was easy to enforce in the past: Anyone who did not use generative AI automatically met the requirements. Today, what matters is whether a provider can use the technology within an approved framework, with defined partners and traceable processes. For Weta, this pressure was the reason to seek out a partner that could move faster.

Moonshine Brings Weta Into Taiwan's Faster AI Environment

Weta found that partner in Moonshine Studio, a Taiwanese animation and VFX company. According to Seah, the collaboration has been underway for around six months. He said Taiwan's development into a national AI hub had also been driven by the island's semiconductor industry. Moonshine had been able to operate without the same client-side restrictions as Weta and had therefore made faster progress.

We spoke with the press on Friday in Taipei after Weta signed an agreement with the Taiwan Creative Content Agency, TAICCA, as part of the Taiwan Creative Content Fest. Under the partnership, TAICCA and Taiwan's Ministry of Culture are expected to identify young talent and introduce it to Weta. Weta could offer internships at its locations in Wellington, Melbourne and Vancouver.

A Taiwanese team is already working in Wellington. According to CTO Kimball Thurston, these artists joined the existing teams before the new agreement and are permanent crew members. Weta currently has no location in Taiwan, according to its own statements. Nevertheless, Seah sees the island as a suitable base for expanding into Asia: Japan, South Korea and Southeast Asia are each three to four hours away from Taipei. Government support, as well as local engineers and storytellers, are additional advantages.

A location in London is also under discussion: Weta recently announced that it plans to open a representative office there. Beyond Taiwan, the company has also sought contacts in Thailand and South Korea. TAICCA chairwoman Sue Wang described the agreement as the largest one Taiwan has concluded with the content industry to date. TAICCA is less focused on an individual location than on enabling the transition from technical hardware to software applications.

Weta Builds Models for Long Films Rather Than Short Clips

The planned AI support is intended to assist artists, not replace them. Weta does not want to replace employees with AI, Seah said. The technology should expand workflows and make them more efficient. According to Thurston, Weta is working on its own models that can alter or create worlds. These models would not fundamentally be built in the pattern of ChatGPT.

At the center is what Thurston calls the “artist layer,” known in the industry as “the harness.” The aim is a tool with which teams can refine stories across the complete timelines of long-form content, rather than merely across “20-second TikTok videos.”

Where customers permit it, Weta also uses external video models such as Kling AI and ByteDance Seedance. Thurston conceded that this is not possible at every studio. At the same time, he said such models have now emerged and have “fundamentally” defeated some of the competition. Because of security requirements, the models must be hosted outside China. ByteDance's global operations make this feasible.

Datasets, Open Standards and Existing Pipelines

18 months ago, Weta launched the open-source Rongotai Model Train Club project, named after a Wellington neighborhood. It documents refinement and LoRA training and records the information used in image metadata. Anyone defining a model must specify which datasets were used. The project builds on the work of the Visual Effects Society's technology committee and MovieLabs.

The technical infrastructure is also part of the plan. Last year, Weta won a Southern Hemisphere Data Center Design Award. The award went to a cooling system that operates without water and is more energy-efficient. Thurston criticized the use of gas turbines to power data centers in this context.

Integration into existing production remains decisive. VFX supervisor Sheldon Stopsack emphasized that the tools must fit Weta's existing pipeline and motion-capture process. The goal is creative control over how the world is designed according to the director's vision. For him and Thurston, the technical possibilities do not end with the synthetic image: People want to see an actor portray a character or creature, even if that character is created with the help of new techniques.

The strongest counterargument to the thesis of a uniform Hollywood position is the state of the available sources. The report cites neither an official industry policy nor a complete list of studios that have changed their rules. Seah's statement is therefore primarily a concrete signal from a major VFX supplier. But it shows that generative AI in Hollywood has already gone from being legally prohibited to becoming a strategic and business decision.

What Actually Follows From the Report

For VFX companies, the practical consequence is clear: Alongside craftsmanship, the ability to present an approvable AI workflow now matters too. This could put smaller providers under pressure if major studios place additional emphasis on AI expertise. For studios, the task in return is to define boundaries for tools, partners and data.

The report does not specify which concrete workflows generative AI is changing in creatures, backgrounds or motion capture. But it shows how far the discussion has already progressed: from the question “Are we allowed to use AI?” to “How do we use it under conditions studios will accept?”

The Variety report provides more context through its conversations with Daniel Seah and Kimball Thurston.