What strikes you first is how real it looks. Two figures grapple atop a crumbling building, the lighting cinematic, the movement fluid, the faces unmistakeable. One of them shouts a line so politically incendiary it would never survive a studio notes session. The other blocks the blows with the practiced ease of a man who has done his own stunts for thirty years. Except neither man agreed to appear in this video. Neither man was there. No camera rolled. No director called action.
The footage, generated by a tool called Seedance 2.0, has circulated widely online in recent weeks, and its arrival has rattled the creative industries in ways that earlier AI video generators simply did not. As first reported by the Sydney Morning Herald, the response from Hollywood's biggest players has been swift and alarmed.
Seedance 2.0 is a generative AI model developed by ByteDance, the Chinese technology company behind TikTok. It is not the first tool of its kind: platforms like Midjourney and OpenAI's Sora have already demonstrated the capacity to produce video from written prompts. But researchers and industry observers say Seedance 2.0 has cleared a quality threshold that earlier tools had not. Dr Justin Harvey, a lecturer in media arts at the University of Technology Sydney, describes its standout feature as an ability to render realistic characters, physical action and human interaction with a fidelity that previous models could not match. He calls earlier AI video outputs, bluntly, "AI slop."
The benchmark used to illustrate the leap in quality has become something of an internet test: a realistic video of Will Smith eating spaghetti. It sounds trivial. It is not. Dr Dana McKay, associate professor in interaction, technology and information at RMIT, explains that the test is genuinely difficult because it demands accurate reproduction of a well-known face performing complex, multi-part physical movements. Seedance 2.0 passed. You can see the result on Instagram and judge for yourself.
Max Schleser, creative arts theme lead at Swinburne's Centre for Transformative Media Technologies, points to another technical advance: visual continuity across shots. Earlier models required creators to generate scenes individually and stitch them together, producing jarring inconsistencies. Seedance 2.0, Schleser says, can maintain consistent characters and settings across a multi-shot sequence, which is the building block of actual storytelling.
The Legal Reckoning
The entertainment industry did not wait long to respond. Major studios including Netflix, Paramount, Warner Bros. Discovery and Disney sent cease-and-desist letters to ByteDance, alleging large-scale copyright infringement. Charles Rivkin, chair of The Motion Picture Association, accused ByteDance of disregarding well-established copyright law, describing the service as one that "operates without meaningful safeguards against infringement."
The US actors' union SAG-AFTRA issued its own condemnation, targeting the tool's unauthorised use of members' voices and likenesses. "This is unacceptable and undercuts the ability of human talent to earn a livelihood," the union said. ByteDance responded by promising to strengthen its content safeguards, though it stopped short of withdrawing the product.
Among the more striking reactions came from Rhett Reese, co-writer of Deadpool and Wolverine, who wrote on X: "I hate to say it. It's likely over for us. In next to no time, one person is going to be able to sit at a computer and create a movie indistinguishable from what Hollywood now releases." His fear is not without logic. If a single person with the creative instincts of a Christopher Nolan can produce a film without a studio, a crew, or a budget, the economics of the entire industry shift.
The Case for Democratisation
But the picture is more complex than a straightforward threat to livelihoods. Some voices, including Mad Max director George Miller, see generative AI as a force for greater equality in filmmaking. Dr McKay puts the argument plainly: big-budget films currently require big budgets. A tool like Seedance 2.0 could, in theory, allow a filmmaker with genuine talent but no financial backing to produce work that looks like it cost tens of millions of dollars.
There are practical upsides too. Dangerous stunt sequences could be generated rather than performed. Multiple takes could be composited into a single seamless scene. Melbourne's ACMI hosted a Disrupt AI Film Festival late last year dedicated to showcasing this emerging form. Internationally, co-writer of Pulp Fiction Roger Avary has partnered with AI studios on three AI-driven productions, and director Darren Aronofsky's studio used AI to produce an American Revolutionary War series.
Dr Harvey is measured in his assessment. He sees AI models as tools that can complement human creativity rather than replace it, much like CGI before them. The film The Brutalist, which won Adrien Brody his second Academy Award, used AI to refine the Hungarian dialogue spoken by its cast. That is a legitimate, careful application of the technology, enhancing human performance rather than substituting for it.
The Harder Questions
Kimberlee Weatherall, co-director of the University of Sydney's Centre for AI, Trust and Governance, focuses on the people most likely to be left behind: the storyboard artists, junior visual effects editors and Foley sound artists who form what she calls the film industry's middle class. Whether new opportunities emerge for them, or whether those roles simply disappear, remains genuinely uncertain. She also raises a structural concern that the optimists tend to skip over. Distribution and marketing remain expensive and complicated, regardless of how cheaply a film can be made. Cinemas have finite capacity. Festival slots are scarce. Generating the film is only one part of getting it seen.
There is a deeper creative risk as well. Generative AI learns from patterns in existing data. Weatherall warns that an industry increasingly reliant on AI tools may find itself producing more of the same, films that converge toward a familiar look and feel because the model has learned what success looks like and amplifies it. Genuine originality, the genuinely unexpected work that redefines a genre, is exactly what pattern-based systems find hardest to produce.
The industry is already hedging. Disney has struck a deal with OpenAI's Sora, and Lionsgate negotiated a similar arrangement with AI research firm Runway. These deals are, in Schleser's view, a pragmatic attempt to keep AI within licensed, controlled ecosystems rather than cede the territory to unregulated tools scraping copyrighted material from the internet.
If there is a lesson here, it is one that resists simple telling. The technology is real, the quality is improving faster than most expected, and the legal and creative questions it raises will not be settled quickly. The honest position is that Seedance 2.0 represents neither the death of cinema nor its liberation. It is a genuinely powerful tool arriving before anyone has agreed on the rules for using it. The most thoughtful voices in this debate, on both sides, seem to understand that the task ahead is not to choose between AI and human creativity, but to work out, carefully and with real stakes in mind, how much of one we are willing to trade for the other.