Netflix says it has used AI on about 300 titles, from planning to visual effects

Netflix has used AI across about 300 titles, its senior director of Asia-Pacific production told a Busan International Film Festival panel. The disclosure, reported by Variety, covered work from planning scenes to visual effects. Alongside him, director Kang Yoon-sung estimated substantial production savings—but warned that technically impressive images can still fall short emotionally.

The panel, “Film Production and AI: New Tools, New Possibilities,” was co-hosted by Netflix and the Motion Picture Association as part of Creative Asia. Netflix production executive Lee Sung-gyun appeared with Kang, director of “The Outlaws,” and Wētā FX chief technology officer Kimball Thurston.

Lee’s examples spanned previsualization—planning how scenes will look before shooting—production design, set-dressing concepts, editing and visual effects. The figure describes titles using AI in various ways, not 300 productions made entirely by AI.

Savings beyond visual effects

Lee screened a clip showing AI and visual effects creating large crowds for the Netflix India title “Glory.” He said budget and schedule constraints would normally have forced those scenes out.

For “The Gulf of Aden,” Kang said calculations prepared for fundraising suggested a production cost reduction of around 40%. That was a projected saving, not a verified final-budget result. Beyond replacing computer-generated imagery and visual effects, AI simplified the shoot and shortened the production period somewhat, he said.

He showed a test clip made entirely with Kling earlier this year, rendered at reduced quality. Kang said cheaper production could revive projects abandoned at the budget stage, opening room for genre and science-fiction films in Korea.

Fewer voices behind the image

Kang said AI can remove the discussions, effort and experience of the many people traditionally involved in filmmaking. The resulting images may reflect one person’s ideas rather than that collective process.

Lee agreed that conveying emotion still requires human intervention. As The Hollywood Reporter reported, his preferred question was not how much AI a title could contain, but what more it could accomplish with AI’s help.

Assistance with boundaries

Thurston offered a narrower example: Wētā’s deep-learning denoiser, a tool for cleaning up image noise. He described it as assistive AI that fixes pixels without being allowed to change creative intent.

• Netflix tests new tools before adopting them and follows strict internal guidelines, Lee said.
• Wētā is building an open-source tool through the Academy Software Foundation to track AI use and provenance—the record of where material comes from, Thurston said.