ByteDance's Seedance 2.0 arrives at Cannes with a 95-minute AI film titled Hell Grind.

      At the 79th Cannes Film Festival, renowned for its abundant cinematic storytelling, one of this year's unforeseen highlights was the innovation brought by AI. On Thursday, ByteDance's cloud service Volcengine introduced its Seedance 2.0 model at Cannes, hosting an AI film showcase and unveiling Hell Grind, a 95-minute feature film generated by AI, claimed to be the world's first full-length AI movie.

      The team behind this feature film hails from the American AI company Higgsfield, with the core video generation model being Seedance 2.0, developed by ByteDance. This production was not a brief AI clip or a proof-of-concept; Hell Grind presented itself as a fully realized, theater-quality narrative feature, demonstrating how swiftly generative AI is transitioning from experimental content to long-form cinematic storytelling.

      What distinguishes this achievement is that long-form video generation has posed one of the greatest technical challenges in AI filmmaking. Current mainstream AI video tools typically generate clips lasting only between 15 and 30 seconds. Creating a feature-length film generally requires integrating tens of thousands of fragmented shots, which often results in inconsistent facial features, unstable scenes, and disrupted visual continuity. Consequently, the output is often unsuitable for professional production processes.

      Seedance 2.0 seems to have overcome many of these obstacles. The film narrates the story of four street kids—Roko, Jaxx, Lulu, and Rein—who discover a mysterious artifact while exploring a museum. This find awakens a sinister force and grants them superpowers, compelling them to unite against the emerging evil while navigating a reality where the lines between truth and illusion begin to blur. After viewing an early version of the film, Chuck Russell reportedly expressed that he genuinely felt empathy for the characters, a rarity in AI-driven cinema.

      The production statistics are equally impressive: the film was completed in just 14 days by a team of 15, with a budget of less than $500,000. In contrast, a traditionally produced film of similar scope could easily incur costs in the tens of millions.

      During the summit, Alex Mashrabov, co-founder of Higgsfield, asserted that the technical foundation for AI-native filmmaking is sufficiently advanced to manifest ambitious cinematic concepts at a much lower cost than conventional production methods. Luc Besson’s SEEN studio is reportedly gearing up to utilize Volcengine’s Seedance 2.0 to create an AI-animated film titled The Furious Five, with Besson attached as the director. This project aims to blend live-action performances with AI generation, eliminating the necessity for motion-capture studios and green screens, thereby allowing everyday shooting environments to directly contribute to the animation process.

      If this holds true, the implications extend beyond mere production efficiency. The existence of a 95-minute AI-generated feature film indicates that narrative-scale generation may no longer be the main limit in filmmaking, shifting the focus toward creative direction rather than budget or team size. For independent creators, this could significantly lower entry barriers and broaden access to feature-length narrative creation.

      However, this shift also raises important structural questions for the film industry. If a feature film can be produced in approximately two weeks at a fraction of conventional costs, mid- and low-tier production workers may face displacement pressures. Moreover, a deeper discussion is emerging regarding authorship: whether the emotional impact generated by AI truly reflects authentic artistic intent or merely the optimization of human response patterns.

      As generative systems grow more adept at delivering coherent and emotionally engaging narratives, the role of human creators might evolve further toward defining intent, taste, and meaning, even as the tools themselves increasingly influence what constitutes effective storytelling.