This year's Cannes International Film Festival is emerging as a global stage for AI-generated imagery.

On May 19, Volcano Engine showcased its AI video generation model Seedance 2.0 at the Cannes Film Palace and hosted an AI imagery summit themed "Ignite Creativity."

The summit attracted Chinese director Jia Zhangke, along with CEOs from top global film and television production companies and advertising agencies, including Luc Besson's SEEN studio and leading international special effects company Outpost VFX.

This was not merely a high-profile roadshow featuring global luminaries but a collaborative exploration of AI's deep applications in theatrical-release filmmaking, content creation, and advertising. After all, many attendees were already heavy users of Seedance2.0.

For instance, Higgsfield relied entirely on the Seedance2.0 model to complete its first 95-minute AI feature film, HELL GRIND, in just 14 days, with plans for a global premiere during the festival. Previously, the company introduced a new marketing video production workflow that supports minute-level generation from product links to finished advertisements, seamlessly adapting to diverse scenarios such as UGC and brand blockbusters, setting a benchmark for industrial-scale AI marketing.

Beyond HELL GRIND, seven Chinese AI-generated films created using Seedance2.0 will also debut at Cannes, including two AI short dramas, The Secret of the Tomb and The Tower of Hunger, which made the official selection. These works stood out among over 1,000 entries from 120 countries, marking China's first AI-produced short dramas in Cannes' official screening section and the only two Chinese entries in that category.

With endorsements from international film and advertising giants and the backing of a top-tier global film festival, Seedance2.0 has expanded its stage from China's Spring Festival Gala to the world.

01 Multimodal Mastery Enables Industrial-Scale Production

With its ability to "generate theatrical-quality blockbusters in seconds," Seedance 2.0 quickly swept through the global tech and film circles upon release, earning accolades such as "surpassing OpenAI's Sora 2" and "marking a singularity moment for AI in filmmaking." Feng Ji, producer of Black Myth: Wukong, hailed it as "the strongest video generation model on Earth, bar none."

To understand Seedance 2.0's breakthrough, one must revisit the previous challenges faced by creators.

Early AI video generation struggled not with image quality but with controllability and coherence. Zhang Ce, a renowned Chinese short film director, once lamented, "Creating a 30-second AI short film requires over 200 attempts, with characters' faces collapsing, costumes changing randomly, and 99% of footage unusable." This uncertainty, dubbed "the card-drawing hell" by the industry, relegated AI generation to a hobbyist stage.

Seedance 2.0 perfectly solved this issue by introducing a dual-branch diffusion transformer architecture. One branch generates high-quality imagery, while the other handles overall narrative and temporal control. Instead of generating frame by frame, the model creates a persistent "world state," ensuring consistent character appearances, actions, spatial relationships, and lighting conditions.

Jia Zhangke described his experience using Seedance 2.0 to create the Lunar New Year film Jia section chief Dance as "astonishing," particularly noting the model's ability to maintain consistent character identities.

Additionally, creators once grappled with making physically plausible scenes, such as water splashing from a shattered glass or a tipped cup. Seedance 2.0 addressed this by incorporating extensive physical motion data during pre-training, ensuring even complex sequences like competitive pair skating—with synchronized jumps, aerial spins, and precise landings—appear natural and coherent.

What truly elevates Seedance 2.0 is its multimodal reference capability. It supports text, image, audio, and video inputs, synchronizing background music, environmental sound effects, and narration with pixel-perfect timing, eliminating the need for post-production editing.

In summary, the model's robust multimodal input, theatrical-grade physical realism, controllability, native stereo audio, and multi-angle storytelling capabilities enable a new technical possibility: industrial-scale production.

Poster for Hell Grind

02 Revolutionizing Efficiency and Cost

At the AI imagery summit, SEEN, a renowned global film production studio, revealed plans to create the first AI-animated feature film, THE FURIOUS FIVE, directed by Luc Besson, known for Léon: The Professional and The Fifth Element.

Global film and television giants are increasingly embracing AI for its unparalleled efficiency gains and cost reductions. Few can resist such a compelling proposition.

Guillaume Lacroix, CEO of SEEN, noted that traditional animation production spans 4–5 years with high costs, whereas the "live-action performance + AI generation" model eliminates the need for professional motion capture studios and green screens, enabling animation creation through everyday filming—a new paradigm for AI-empowered filmmaking.

Duncan McWilliam, founder of Outpost VFX, a leading global special effects company, echoed this sentiment. Traditional filmmaking follows a linear process, with each department executing based on its interpretation, leading to information loss.

Seedance2.0, however, synchronizes all stages, allowing departments to visualize effects before filming, align creative goals, adjust ideas swiftly, and mitigate project risks. "The film industry has long lacked a shared blueprint where everyone sees the same final result," McWilliam said. "That piece just arrived."

The upcoming global premiere of HELL GRIND exemplifies this. The 95-minute theatrical film, created using Seedance2.0 by a 15-person team in 14 days, cost under $500,000, compared to the $50 million typically required for a traditional production of similar scale.

Notably, several global branding giants have integrated Seedance 2.0 into their content production workflows. For instance, WPP CTO and Havas's Prose on Pixels will leverage it to generate scalable creative assets. Spain's AI creative platform Magnific has also adopted Seedance2.0 to provide one-stop content production services globally.

03 When AI Becomes "Infrastructure"

At the summit, Tan Dai, President of Volcano Engine, remarked, "AI has solved issues of creative efficiency and technical barriers. It's not replacing creation but liberating creators from tedious execution, allowing them to focus on character development and storytelling, returning the industry to its creative essence."

His comments addressed concerns about AI's impact on human creators. If AI can generate theatrical films at the click of a button, where does that leave human artists?

Yet, such scenarios are not new in film history.

When synchronized sound recording emerged in 1927, transitioning cinema from silent to sound, some feared it would "destroy the purity of film's physical art." However, it enriched narrative depth. Similarly, the advent of Technicolor in 1935 raised concerns about "undermining film's atmospheric quality," but it soon expanded visual expression.

Every technological leap initially faces skepticism or resistance but is eventually embraced.

Director Jia Zhangke is among the first to adopt AI, viewing it as a "technological extension of Image Writing (image writing)" that revolutionizes traditional filmmaking centered on cameras and recorders, offering more accessible tools for image creation. "It's like writers transitioning from pen and paper to computers," he said. "Ordinary people can now generate their imagined worlds with AI, further enriching creative possibilities."

Gong Li, chair of the World AI Film Festival (WAIFF), also supports AI's application, stating, "I've always believed that film's warmth comes from the human heart, while technology enhances our freedom and depth of expression."

Today, video AI, exemplified by Seedance 2.0, is deeply integrating with global applications in film, advertising, e-commerce, gaming, and digital humans. In the future, it will become as fundamental as water, electricity, and coal—a reliable public utility for image production. At the pinnacle of this chain lie the unshakable elements technology can never replace: human creativity, emotion, and experience.