Hollywood studios are scrambling to respond to a new Chinese artificial intelligence video model that can produce polished, film-like scenes from a handful of written instructions, raising fresh questions about copyright, performers’ rights and the economics of screen production.
The system, called Seedance 2.0, was developed by ByteDance, the owner of TikTok. In company materials, ByteDance describes Seedance 2.0 as an “audio-video joint generation” model designed for “industrial-grade creation scenarios”, with improved realism and control over motion, lighting and camera movement.
What has unsettled the entertainment business is not the existence of another text-to-video tool, but how quickly Seedance clips began circulating online that appeared to mimic the look and cadence of professional studio work. Reports this week described hyper-realistic sequences featuring well-known actors’ likenesses and recognisable characters from major franchises, alongside synthetic dialogue and sound design. Industry groups and studios argue that such outputs are unlikely to be possible without training data that includes protected material and without the consent of performers whose faces and voices may be replicated.
The backlash has moved rapidly from social media to legal letters. Trade publication reports said Disney and Paramount sent ByteDance cease-and-desist notices alleging infringement tied to Seedance-generated clips. Business reporting published on 18 February said Netflix also sent a legal notice, characterising the tool as enabling knock-offs of its shows and characters. ByteDance has said it respects intellectual property rights and is strengthening safeguards.
Seedance’s timing is part of the story. The ByteDance “Seed” team’s own posts indicate Seedance 2.0 is a step-change over earlier versions, positioning it as a creator product with director-like controls and higher usability for complex motion. Even in a market already crowded with generative tools, the arrival of higher-fidelity video has sharpened a long-running dispute: whether generative AI should be allowed to learn from copyrighted films, television, journalism, music and art without explicit licences, and whether outputs that imitate a studio’s characters, settings and signature style should be treated as derivative works.
For Hollywood, the immediate concern is enforcement. Copyright litigation is slow, often cross-border, and fact-intensive. A further complication is that Seedance 2.0 has been reported as being available only in China in its current form, while the clips drawing attention may have been shared globally through users and re-uploads. Studios can demand takedowns and threaten claims, but policing thousands of short clips, prompts and reposts is closer to content moderation than conventional rights management.
The next layer is labour. Screen actors and writers have already fought battles over AI protections and consent, and the new generation of models raises a practical question: how much of a performance, and how much of a production pipeline, can be simulated. The Associated Press reported criticism from the Motion Picture Association and SAG-AFTRA, focusing on alleged misuse of copyrighted works and unauthorised likenesses. If the model can reliably assemble convincing action sequences, voice lines and ambient sound, it could pull work away from visual effects houses, sound teams, and even parts of pre-visualisation and editing.
Yet the same capabilities that alarm major studios are attractive to smaller producers and advertisers. High-quality generative video could reduce the cost of producing short-form dramas and promotional content, particularly in markets where budgets are small and turnaround times are tight. In that scenario, the competitive pressure is less about replacing cinema releases and more about flooding platforms with low-cost, plausible-looking video in genres that previously required heavy effects spending.
Seedance has also become a proxy for a wider geopolitical debate about who sets the rules for generative AI. Axios described the model as part of a broader challenge from Chinese AI firms, with US studios leaning on copyright law while Chinese developers push rapidly on capability. ByteDance’s launch adds to a pattern in which major Chinese companies present frontier-level systems while governments in the US, Europe and Asia argue about disclosure, watermarking, licensing and liability.
What happens next will likely be determined less by a single tool than by whether markets converge on enforceable standards. Studios want mechanisms that prevent models generating protected characters and that compensate rights holders when their catalogues are used in training. Developers, including ByteDance, say they are building safeguards, but the details—what data was used, how it is filtered, and how rights are cleared—are not publicly settled.
For Hollywood, Seedance 2.0 has accelerated a shift already under way: the contest is no longer whether AI can make video, but how close it can get to professional output, and which legal and commercial frameworks will govern the raw material of modern culture.
