Chinese AI floods world with mass-produced films

Full text

“A 1960s American retro-futurist city street. A zombie outbreak has just erupted. The city has freshly emerged from battle, wearing the stillness that follows combat. It’s high summer, the sun directly overhead. Cold War, space-age streamlining: rounded, science-fiction domes. Chaos everywhere — wrecked vintage cars, scattered newspapers, blood on the pavement, storefronts from another decade. Zombie corpses litter the street. The leisure of the setting sits in violent contrast to the dead.”

The scene is stitched from so many motifs of the old Hollywood zombie picture that any cinephile would feel a flush of recognition. It’s a whimsical love story between a cowboy robot and a mannequin, and the quality of its storytelling and camerawork rivals that of a Netflix or HBO production. The entire three-minute movie is the work of a 28-year-old wedding photographer from a small rural town in Yunnan Province — a man who has never once left China. He goes by Mx-Shell online, and the paragraph you read a moment ago is a slice of the prompt he fed to AI.

Today, studios produce micro-dramas the way automotive companies build cars: on an assembly line tuned for sheer volume.

Working with Seedance 2.0 — a video generation tool developed by Chinese tech company ByteDance — a handful of asset images, and a script he wrote himself, Mx-Shell produced Zombie Scavenger in just 10 days, spending around 3,000 yuan in token costs (approximately $442). On Douyin, China’s version of TikTok, it has been watched more than 70 million times. Overseas, it has picked up another 12 or 13 million views. PJ Ace, an American AI film producer, went looking for Mx-Shell online to offer him a job. Back in China, the state-run newspaper Wen Wei Po wrote that the creator’s lack of credentials — “the disadvantage that would have shut anyone else out of any other era of filmmaking” — had been erased entirely by AI. He had broken free of the old constraints of technology, capital, and connections to become what the paper called an “army of one.”

Everyone I met in China’s nascent AI film industry this past May told me this story with exuberance and starry eyes. He only started learning AI in January! He had no experience before! Mx-Shell’s success had already hardened into the industry’s founding myth — proof that was passed around every dinner table. In the age of AI, anything is possible. Just 3,000 yuan can create a cinematic miracle.

The content is made for the thumb and the phone, with plots calibrated to release dopamine on cue.

From the outside, China looks like it is leading the world in AI-generated video, cracking Hollywood’s code. Some tech analysts even believe AI micro-dramas could be strengthening China’s soft power (a point I carefully disagree with). Bundle Mx-Shell’s story with the arrival of models like Seedance 2.0 and Kling, a tool developed by Beijing-based tech company Kuaishou, and you might conclude this is another DeepSeek moment, another example of China winning.

As a movie lover, Zombie Scavenger made me want to find the second Mx-Shell. But when I set out to write more triumphant stories about China’s success with AI-generated video — about Chinese creators being quicker to embrace AI, about Chinese-developed tools unleashing people’s creativity, and about Chinese AI content winning Westerners’ hearts and minds — I found that Zombie Scavenger and Mx-Shell were an anomaly. The situation I discovered was grim: In China, advanced models, short-video platforms, cloud infrastructure, AI-eager workers, micro-drama studios, and permissive state support have converged to create a monstrous Industrial-Slop Complex, and it has nothing to do with tech diffusion, innovation, or even entertainment.

The ubiquitous micro-drama

Slop, low-quality digital content that can be produced quickly and in large quantities with AI, was so prevalent in 2025 that Merriam-Webster declared it the word of the year.

In China, much of it is built atop the nation’s micro-drama industry, a strange yet profitable juggernaut that churns out cheap, fast-paced videos also known as “short dramas” or “verticals” (because they’re produced in the 9:16 aspect ratio). The content is made for the thumb and the phone, with plots calibrated to release dopamine on cue. Some hail the form as an original Chinese invention — an entertainment current running from China to the West. Today, studios produce micro-dramas the way automotive companies build cars: on an assembly line tuned for sheer volume. Western users can watch them on ReelShort, DramaBox, and FlexTV, Chinese-backed apps built for a single end: making content go viral to eat people’s attention.

Every character had the same glossy, overlit skin and stiff gait.

A May 2026 report from Yale University argues that Chinese micro-dramas transform “dramaturgy into conversion architecture” — they essentially exploit addictive dynamics to push people to pay. It notes that older serial forms allowed for “ambiguity, scenes that do not convert, moments that deepen character development,” whereas micro-dramas demand instant payoff: “Viewers pay for escalation on schedule, even as the system is designed to defer closure indefinitely.”

Micro-dramas can be produced strikingly fast and cheaply, and the production pipeline is often controlled by the platforms themselves, which also collect and use audience data. The line is now being reassembled by AI , which is enabling the micro-drama industry to generate videos at a velocity with no real precedent. The New York Times reported that creators uploaded nearly 50,000 new AI-generated micro-dramas to Douyin in March 2026 alone. That’s almost enough to match the platform’s entire 2025 output.

These dramas live in a place with no bearing on human artistic experience. They are monstrous feats of psychological hacking.

According to China’s Netcasting Services Association, the nation’s short-drama industry produced about 128,000 micro-dramas in the first quarter of 2026, and 95% of them were fully AI-generated. One industry forecast predicts five million will be produced in 2027 — a hundredfold expansion over 2025’s figure. At a conservative two minutes per drama, 5 million dramas total 10 million minutes of entertainment. It would take someone watching two hours a day roughly 230 years to get through it all.

The boom is visceral enough that one AI filmmaker told me that, when she opened a job app, she found more than 400 companies hiring for AI short-drama positions in her area alone.

I didn’t have 230 years, but I did have a few days to immerse myself in AI-generated micro-dramas, the majority of which are animated. Instead of whimsical settings like Zombie Scavenger’s “American retro-futurist city,” I discovered countless retellings of just a handful of scripts in just a few genres (neo-traditional Chinese fantasy, time-travel revenge, cut-rate science fiction, etc.), each rendered in a predictable palette with a cheesy plot and over-the-top dialogue.

When I moved on to watch contest winners, I found slightly more original storylines, but little visual improvement. Every character had the same glossy, overlit skin and stiff gait. Each setting, whether a Song dynasty court or an intergalactic space station, was overbuilt and overfurnished and way too jelly-like, luxuriant to the point of vertigo. Somewhere around the 15th film, my early excitement about AI-generated cinema was completely gone, replaced by a dizziness and faint nausea from being trapped in the uncanny valley for too long with beings that imitated life but contained none of its friction.

Rui Pu

The incentives behind the slop

I should admit here that I am a techno-optimist. In early 2025, when the AI-generated clip of Sam Altman stealing GPUs went viral, I laughed hard, convinced this would be great.

But almost none of the AI micro-dramas being made today were created in pursuit of anything artistic or because someone loved the idea that spawned them. These dramas live in a place with no bearing on human artistic experience. They are monstrous feats of psychological hacking, and they exist because of a condition that philosopher C. Thi Nguyen calls “value capture” in his most recent book, The Score : “when we are exposed to a simple public scoring system — a ranking, a metric, a numerical score — and it takes over our decision-making.”

Longer and longer hours, lower and lower pay, and beneath the complaints about wages, the knowledge that no one is watching.

Because Beijing has sent out signals that it sees value in AI-generated content (AIGC), tech companies and local governments — many of which are both risk-averse and eager to gain favor with the central government — are simply racing to outdo one another in the relatively low-stakes arena.

For governments, this means sponsoring competitions and hosting award shows to publicize winners. In April 2026, Hubei’s provincial broadcasting bureau announced the Audiovisual Yangtze , an AIGC short-film and short-drama competition, under the slogan “AI Empowers, Wisdom Gathers at the Yangtze.” All entries had to be at least 80% AI-generated. In 2024, Shenzhen’s Longgang district staged its first annual Visual Arts Created by AI Technology (VACAT) competition. The competition is co-hosted by the district government, Shanghai Film Co., and Bilibili, China’s version of YouTube, and pitched as “the Oscars of AI visual art.” Mentougou, the district on Beijing’s western edge — once known for being both the capital’s poorest district and the seam of its richest coal deposits — rolled out talent-support measures for “ AI+ premium micro-dramas ” in June 2026. Shenzhen launched an Intelligence-Creation Cup for AI-animated dramas earlier in the year and published alongside it the country’s first vocational training standard for AI drama production.

This government interest sends its own signal to tech companies, who respond by pushing employees to create more AIGC, faster, in the hope of producing viral clips that grab attention and generate revenue.

Yuan describes her job in honest language that is hard to forget: She produces shit at her desk.

In April 2026, Nanfengchuang , one of China’s largest economic news magazines, published an account of the gruesome production process of an AI micro-drama, told from the perspective of a recent graduate it calls “Yuan.” Her employer is a young AI-video company in Nanjing, and its production method is simple and closed-loop: Someone downloads a viral video, extracts its dialogue, and hands the transcript to Yuan. The plots are usually about time-travel revenge, betrayal by an ex-husband, a biological daughter frozen out by a stepchild, etc. Yuan reproduces something close enough to the original. Her quota is 10 drama series a month, each around 30 minutes. The first few episodes are released for free, and the rest are unlocked if the viewer pays or sits through an ad. Yuan describes her job in honest language that is hard to forget: She produces shit at her desk.

A report written by business journalist Chen Zijie and shared on WeChat the same month recounted how workers have to queue for tokens to get access to the popular Seedance 2.0 video generation model. “You might stay logged in from the start of work to the end of the workday and still not get a spot,” wrote Chen. “So working overtime until the early hours has become the norm; sometimes you even have to crawl out of bed at two or three in the morning to work ‘off-peak.’” The report also highlighted the lengths companies will go to in order to produce the few dramas that do go viral: “The scripts get more and more sensational — vulgar scenes like child abuse and beating pregnant women crop up one after another. Some companies even stipulate that if the opening doesn’t contain ‘borderline’ provocative content, it will be rejected outright.”

The content ghost cities

These stories, and I am certain there are many more, show China’s Industrial-Slop Complex to be both exploitative and self-defeating. The workers caught inside it will bring to mind what German philosopher Karl Marx set down in his Economic and Philosophic Manuscripts of 1844 : that under such a mode of production, human nature itself is alienated and worn away.

I would argue those stories prove something further: that the mass manufacture of AI slop, and the celebrated “China model” it is said to represent, is not the critical lever by which an economy grows over the long run. According to reporting by Securities Daily , one of China’s four major securities newspapers, just 0.16% of AI-generated short dramas are “hits,” meaning they generate more than 100 million views. One exec told the paper that 90% of the companies making AIGC are losing money.

The Industrial-Slop Complex narrows our choices by feeding us ‘what we want,’ but not […] ‘what we want to want.’

It seems those inside China’s AIGC industry are coming to the same conclusion. Discussions on the social networking platform Xiaohongshu suggest that the era of the “AI micro-drama sweatshop” may already have reached its end. Longer and longer hours, lower and lower pay, and beneath the complaints about wages, the knowledge that no one is watching.

My search into AI slop in China has been a journey of alternating surprise and disappointment. In that mood, I called Jamie Luk , a friend of a different generation. Restless, electric, and now in his early 70s, Luk came up in Hong Kong cinema’s Zhang Che era and worked alongside Tsui Hark and Jackie Chan in the ‘70s, ‘80s, and ‘90s. Today, he’s in Shenzhen trying to make his first AI feature — an actual “AI movie,” with a budget of 8 million RMB (roughly $1.2 million). “Don’t waste your time on micro-dramas,” he kept telling me. “Don’t waste your time. They ruined the film industry.” By “they,” he meant the Industrial-Slop Complex, which narrows our choices by feeding us “what we want,” but not what the late philosopher Harry Frankfurt called “what we want to want.”

The mass manufacture of AI slop is not the critical lever by which an economy grows over the long run.

Luk’s most active years coincided with Hong Kong cinema’s glittering golden age (an era that also mass-produced plenty of human-made slop). Our friendship began with a brief, dreamlike internship at a Hong Kong production company during my first summer break from college, when I was 19 and hungry for the world of film. Hong Kong cinema imprinted itself on me in textures that felt magical and contradictions that felt impossible: the rough theatricality of Shaw Brothers sets, the blunt moral certainty of the heroes, the crude sound effects, Jackie Chan’s physical resilience, the cheap B-movies that were somehow — mysteriously — sincere.

About a decade after Hong Kong’s cinematic boom, China’s real estate industry experienced a boom of its own. Driven by memetic competition among local governments and incentives pointed in the wrong direction — just like the Industrial-Slop Complex — it threw up dozens of near-identical towns across the country. But no one came to populate them. Thousands of apartments sat empty. Vacant office towers peppered the skyline, with whole floors of glass where the lights were never once switched on.

My disenchantment with AI micro-dramas has me thinking about those ghost cities. Eventually, the real estate industry moved on from them, and the abandoned buildings became overgrown with weeds. I’m counting down the days until the same happens to China’s AI slop.

This article China is mass-producing slop at a scale the world has never seen is featured on Big Think .