China’s first fully AI-generated movie has flopped, earning just 1.7 million yuan (£191,000) at the box office despite costing “tens of millions” to make, according to its producer.
Sanxingdui: Future Past tells a confusing tale of a future society in which every decision has been outsourced to algorithms and AI, leading to a global catastrophe. The only way to avert doom is to unlock a hidden code in the Sanxingdui bronze sculptures, linked to the ancient kingdom of Shu. Cue astronauts racing through ancient Egypt and archaeologists diving through space-time vortices.
All of the film’s visuals and some of its sound effects were created using AI, while it was edited by humans.
The Quality Problem
The Guardian watched the film in a virtually empty cinema in Beijing and found it no worse than plenty of human-made action flicks — though perhaps a bit worse.
Many of the characters were uncanny — life-like for the most part, and uncanny valley for the remainder. The Chinese characters seemed more realistic than the white people. All of them struggled with running, falling into the steady, smooth gait of video-game characters. One main character, an astronaut named Wu Xingyan, looks creepily like Tilly Norwood, the controversial “AI actor” appearing in an upcoming Hollywood film.
The use of the production company’s in-house AI scriptwriting tool was evident in lines like: “That is the superbrain superintelligence that costs $10 million a month to subscribe to.”
Why It Flopped
Sanxingdui flopped because “the movie isn’t entertaining,” said Yu Yaqin, an independent film critic in Beijing. But at least in the short term, “AI films will continue to be produced in China because investors place a high value on AI,” she said.
One person who works in screenplay development in Beijing said the majority of projects she has been approached by in recent months have been for AI films. “Chinese investors believe this is the future,” she said, not least because AI films could be much cheaper. “What used to take 100 people to do, now can be done by five people, and the rest won’t have any jobs.”
The Microdrama Boom
Chinese audiences are no stranger to AI-generated content. Tens of thousands of AI-generated microdramas — one- to three-minute dramas designed for mobile phones — are uploaded to streaming platforms each month. The industry, which didn’t exist a few years ago, is now worth an estimated $3 billion.
Sanxingdui’s poor performance may also be because people don’t want to pay 50 yuan for a cinema ticket when they can watch similar content on their phone for free. “AI-generated microdramas are selling a way to kill time, while films in cinemas are selling an experience worth spending time on,” wrote one film blogger.
The Bottom Line
China’s first fully AI-generated film flopped at the box office, earning just £191,000 despite costing tens of millions. Critics say the movie simply isn’t entertaining. But investors remain convinced AI is the future of cinema, and the AI microdrama industry is already worth $3 billion. The question is whether audiences will ever pay cinema prices for content they can watch for free on their phones.
My Opinion
The flop of Sanxingdui: Future Past is an example of what happens when technology outpaces storytelling. The film did not fail because the AI was bad. It rather failed because the script was bad, the plot was incoherent, and nobody involved asked whether the movie was worth watching. The fact that it cost tens of millions and earned £191,000 also points to it being a product problem.
Here is what I find most revealing. The investors who funded this film did not care whether it was good. They cared that it was AI. One screenplay developer said the majority of projects she has been approached about recently have been AI films because “Chinese investors believe this is the future.”
The microdrama comparison is also worth looking at. AI-generated microdramas are worth $3 billion because they are short, cheap, and designed to be consumed on a phone. Especially because they do not compete with cinema or compete with scrolling. The moment AI tries to compete with a two-hour theatrical experience, the limitations become obvious. You cannot fake the craft of a good film with a scriptwriting tool called Bowen.
The takeaway from all this is not whether AI can make a film. It clearly can. The question is whether anyone will pay to watch it. As it stands, the answer is no, and until investors understand that technology is not a substitute for talent, they will keep losing tens of millions on films nobody wants to see. The future of cinema may be AI-assisted. But it will never be AI-only.
