Generative AI has already changed commercial advertising, is replacing countless processes within film and TV production and is poised to upend even newer technology such as virtual production, says filmmakers and technologists who have spoken to Variety on the sidelines of FilMart.

With significant presence from foundational model builders on the market floor, to AI apps targeting every niche filmmaking use and numerous seminars promoting links, however tenuous, to AI, the conversation at FilMart has firmly shifted from early adoption to “directable AI.”

Kling AI, the foundational model released by China-listed Kuaishou, has set a series of exhibitions and deep-dive workshops showcasing its latest 3.0 model. The update marks a significant leap in “directable AI,” a suite of tools designed to give filmmakers granular control over movement and composition that was previously the domain of expensive VFX houses.

Its new 3.0 model, released in February, introduces advanced camera control and character consistency that will enable users to direct the performance with greater exactitude.

The impact of these tools is already visible in high-end Chinese productions. The historical drama “Swords Into Plowshares,” a series shown on national broadcaster CCTV1, serves as a primary case study for these shifting workflows.

According to Chen Yi, founder of “Swords” VFX provider Timeaxis Studios, Kling AI was integrated into every stage of the pipeline, from rapidly generated pre-vis material to generating final effects plates for compositing.

In one instance, the team generated a picture of a scavenging raven, and animated it entirely within Kling AI. It was then composited onto live-action background plates.

“AI-enhanced workflows proved to be three to four times more efficient than traditional CG,” said Chen. “Once current limitations in resolution, color encoding and color gamut are fully resolved, efficiency could skyrocket by eight to 10 times.”

That efficiency is also driving vast changes in the commercial world in Asia. According to cult Singaporean film director Gavin Lim (“Diamond Dogs”), generative AI is already cannibalizing the commercials sector.

“We’re using it for 3D, generating architectural renderings, simulating time lapses. Almost no advertising content now goes beyond one month. So actually this AI fits perfectly because advertising is disposable,” said Lim.

He contends that generative AI will become even more prevalent when brands start to micro-target consumers with customised advertising.

Lim founded the Singapore educational collective AI Film Lab and is firmly outspoken about how this technology levels the playing field.

“Hollywood is scared,” Lim asserts. “Gen AI is going to help small, less-capitalized film industries in Asia compete with multi-million dollar budgets on a visual scale.”

This extends to the use of Asian foundational models. But Lim has faced limitations in U.S. foundational models while training other filmmakers through his AI Film Lab collective.

“American AI is too fussy and, frankly, it doesn’t handle Asian faces well,” Lim explains. He recounts frequent issues with U.S.-based platforms triggering safety guardrails by mischaracterizing Asian adult faces as “underaged,” even when the content generated is in no way sexualized.

“It’s not that the Chinese ones like Kling are not safe. They have a lot of guardrails, you can’t just generate a bikini shot, or whatever. But it’s not ridiculous like [Western models], who just see an Asian face and they say it’s underage.”

The upheaval in jobs however, is real. A representative from Kling AI predicts the emergence of entirely new roles: AI VFX artists and AI cinematographers. Tedious tasks like rotoscoping and storyboarding are being automated out of existence, making space for higher-level creative decision-making.

Lim thinks that gen-AI may spell disaster even for newer technologies like LED virtual production and in-camera VFX. New tools coming online which streamline and automate greenscreen keying, color correction and matching of scenes, and digital background asset generation may support his thesis.

“Stop playing with this LED technology, it will die,” said Lim. “3D guys are now starting their own AI companies.”

However, the human element remains the ultimate friction point. Lim is adamant about maintaining a “human in the loop,” stating he is fundamentally against any system that removes the director’s final intent.

This sentiment is echoed by producer Tan Bee Thiam, who views AI as a beginning rather than an end.

While his team uses generative platforms for concept art and visualizing complex ideas during development, he stresses that it remains a tool for exploration. “We use it as a starting point rather than the end. That would take away the joy of making the film.”

Ultimately, AI’s downfall might come only from the laziness of its users. When rival foundational model builder Seedance released its latest model, the internet was flooded with videos of choreographed fight scenes between well-known actors like Brad Pitt and Tom Cruise.

According to Gavin Lim, it is this “AI slop” that must be combatted.

“Because AI filmmaking is so easy, most users are just playing with that thing. But filmmaking is not easy, there’s a lot of deliberation, and that person who deliberates and keeps going at it for six months, makes it good,” said Lim.

“How do we fight slop? How do we legitimize our work? How do we respect ourselves? Don’t touch any celebrities. Don’t touch existing IP. Don’t go for the cheap joke.”