11/18 2026
407
Which type of AI application, then, is consuming the most computing power and attention today?
A recently circulated research report offers an answer: of all Token consumption in China, AI short dramas and video generation account for 55%, e-commerce and marketing 24%, and software development a mere 15%. Meanwhile, at the Busan Film Festival, directors of both long-form and short-form productions have begun experimenting with AI creation—and 95% of the latest short dramas released have involved AI in one form or another.
"Even if it's shot with real people, the visual effects and special effects are definitely done by AI." This has become the consensus among many film and television practitioners.
AI video appears to be entering an era of mass adoption, emerging as the form that commands the most attention.
And not just attention. Some analysts argue that if you survey AI applications across industries over the past two years one by one, film and television is likely to be the first category to achieve genuine profitability.
On one hand, AI short dramas are gaining ground at a visible pace; on the other, the traditional film and television industry's rear guard—costumes, makeup, and props—is losing orders en masse. How exactly does AI slash production costs so decisively and pry open the door to profitability? And beneath this "massacre," how many practitioners are being swept up and replaced?
On the first day back at work after the 2026 Spring Festival, short drama producer Xiao Mo opened his phone as usual, waiting for a notice to join a production. The notice never came. He later learned that the leading short drama production company in Xi'an where he worked would not start any live-action short drama projects until at least the end of March, and was instead going "all in on AI." No layoff announcement, no N+1 compensation—just no more work. Half a year earlier, a producer like him could earn up to 1,600 yuan a day; nearly a month after the Spring Festival, his income stood at zero.
Xiao Mo is not an isolated case. Over the past two years, AI short dramas have been frantically crowding out the supply in the short drama track on short video platforms. According to monitoring by DataEye and the China Netcasting Services Association, Douyin alone added 221,900 AI short dramas in the first half of the year, with a peak average of one new AI short drama going online every 36 seconds. By August, AI short dramas had claimed seven of the top ten spots on Douyin's short drama popularity rankings.
With the iteration of AI simulated human technology, the single-drama views of top hit AI dramas have reached the 2 billion level, and the overall deduplicated user base of short dramas has reached 851 million—roughly two out of every three internet users are scrolling through short dramas.
To understand why AI short dramas are so aggressive and can capture user attention so quickly, one must first look at how low their costs have fallen.
Even the most rudimentary traditional short drama requires initial funding of two to three million yuan. And that's with fairly ordinary costumes, makeup, and props. Where does the money go? Actors' fees, extras' wages, venue rental, lighting equipment, camera crews—every item is a hard expense. A short drama production team, even the leanest one, typically needs twenty to thirty people. Moreover, live-action crews have long cycles and extremely high trial-and-error costs. If you shoot something wrong, you have to reshoot, and reshoots require re-coordinating venues, personnel, and schedules—pull one thread and the whole thing moves.
It can be said that traditional film and television is "people serving the process." Turning a script into images requires going through countless stages—financing, team building, venue coordination, post-production—and each stage supports a large amount of employment. AI short dramas crush all of this.
AI short dramas have become "the process serving people." In theory, as long as you can write prompts and have enough computing power, one person can do the work of an entire crew. Xi'an Fenghang once produced nearly a hundred live-action dramas a month; now its output has shrunk to single digits. Zhengzhou Huayang Nianhua's live-action projects have been cut by 30%.
Someone has done the math: a short drama that once cost millions can be produced with AI for 100,000 to 200,000 yuan, covering the entire production cost. One editor and one screenwriter with two or three prompt writers can finish in less than a week what used to take dozens of people several months. AI's trial-and-error capability is virtually unlimited—if the special effects aren't satisfactory or location scouting is difficult, just change the prompt and parameters, which is far more cost-effective than finding venues and organizing actors.
Market data shows that script purchase prices have dropped from 30,000 yuan to 10,000 yuan, voice acting has been compressed from 140 yuan/hour to 45 yuan, and extras' daily wages have been halved from 150 yuan to 80 yuan. Many crews only value the male and female leads; everything else can be replaced by AI.
AI has also penetrated from the periphery, permeating the entire short drama industry.
If you look only at job titles, actors and extras seem to be the hardest hit. Two years ago, short dramas were still regarded by many as a lifeline in the film and television winter; veteran actors in Hengdian said everyone "was saved by short dramas." Large numbers of film and television graduates who couldn't get into big productions flooded into short dramas, and Xiao Qian was one of them. Seventy percent of her classmates entered the short drama field because "the money comes fast"—assistant directors' daily wages could rise from 200 yuan to 800 yuan within half a year.
But the arrival of AI changed all of that. In the film and television bases of Hengdian, Zhengzhou, and Xi'an, mid-tier actors' orders have shifted to AI short dramas; some can't get work even after taking a one-third pay cut, and after extras' daily wages were halved, some changed careers and went home to farm.
This is painful, of course, but the plight of actors can at least be seen. The truly silent massacre is happening to those job types for whom even "being seen" is difficult.
Over the past few years, the vast majority of practitioners in the film and television industry have actually made their living around short dramas. A large number of post-production editors, lighting technicians, photographers, camera operators, the various extras and set builders in Hengdian, and the people in film and television studios who build sets, manage costumes and props, and even manage boxed lunches... the main thing they serve is short dramas. These positions don't have the glamour of stars and don't trend on social media, but they form the true foundation of the entire industry.
Once a short drama is finished, it supports a silent industrial chain from pre-production to post-production, from the front of the camera to the boxed lunches.
On-set producers, assistant directors, casting coordinators, lighting technicians, grips... their labor is bound to the physical space of "the set." As long as there is no live-action shooting, they have nothing to do.
After AI-ification, this foundation has collapsed. A short drama crew that once needed twenty to thirty people and ordered a dozen boxed lunches has now shrunk to three or four people who finish all the special effects in a single server room. In Hengdian, even the boxed lunch business has gone quiet.
This is precisely why the film and television industry is the loudest in criticizing AI... it has essentially stolen the livelihood of an entire rear guard. Lighting technicians, camera operators, extras, costume and prop staff, grips—every link in the chain is being compressed, erased, until only the prompt writers, the director, and the male and female leads remain.
Why short dramas? Because the vast majority of people in the film and television industry were never centered around those few big-budget productions. Those top-tier projects—costume fantasy, xianxia, historical dramas with budgets in the hundreds of millions—number only a few per year, and most practitioners never get such work. The low barrier to entry of short dramas is precisely what makes them the fastest to be replaced; AI is better than they are at creating cheap visual spectacles, better at catering to the fragmented, scroll-heavy tastes, and its costs are rock bottom.
Since AI has already replaced the rear guard of film and television production and halved costs, are production crews making money?
AI's aggressive advance has completely shattered the short drama get-rich dream. But can short dramas get rich by cutting large numbers of behind-the-scenes workers?
When AI short dramas first emerged in 2025, revenue sharing per 10,000 views could reach 30 to 100 yuan, and stories of an 8,000-yuan project earning over 500,000 yuan in revenue sharing did happen. At that time, platforms were competing for content, subsidies were generous, and computing power was cheap—the computing cost for one episode of an AI short drama was only a dozen yuan or so. But prices built up by subsidies are greenhouse prices. After large numbers of people rushed in, subsidies were withdrawn, computing power returned to real market prices, and per-view revenue sharing began to fall. By November 2026, revenue sharing per 10,000 views had dropped to 5 to 10 yuan, with tail-end assembly-line works getting only 1 to 2 yuan. Meanwhile, computing costs rose from a dozen yuan per episode to eighty or ninety yuan. Input and output—profit margins are being squeezed from both ends.
As the technical barrier lowers and the number of short dramas grows, hit AI short dramas are becoming rarer, and the vast majority are operating at a loss.
Of the more than 200,000 AI short dramas launched on Douyin in the first half of 2026, only 1,055 surpassed 100 million views, accounting for less than 0.5%; about 2,900 reached the break-even line, accounting for 1.3%. On average, only one out of every 77 new dramas can recoup its investment. The platform's revenue-sharing pool is only so big, and with more than ten times as many people dividing it, each person's share is naturally diluted. What subsidies can cover are works that can at least reach several million views. A large number of AI short dramas are stuck at hundreds of thousands or even tens of thousands of views, making it difficult even to trigger the revenue-sharing threshold.
So whose money are those 1.3% who recoup their investment making?
The answer is the existing stock taken from the traditional film and television industry. With lower costs, it has taken over the audience, the traffic, and the platform revenue sharing that originally belonged to live-action short dramas.
The platform's revenue-sharing pool has not expanded because of the emergence of AI short dramas; it is largely still the same pie. The time viewers spend on short dramas, the procurement budgets platforms are willing to pay for content, and the budgets advertisers allocate to short dramas have not doubled within a year. The traffic and revenue sharing that AI short dramas have squeezed in were cut from the share that originally belonged to live-action short dramas.
In other words, even if AI short dramas themselves operate at a loss, their profitability is still built on the massacre of their peers. A niche market that once needed dozens of people and millions of yuan in costs to sustain has been forcibly occupied by three or four people and 200,000 yuan in costs.
Once the profit logic of AI short dramas is clear, the predicament of the film and television industry is not hard to understand. The meat that AI teams have snatched away looks like a victory of cost advantage on a single-project basis, but viewed across the entire industry, it means that an entire employment chain dependent on live-action shooting is contracting. The extent of the contraction is even greater than what the short drama industry itself feels.
Short dramas are just the first link to be affected, because they are small in scale, short in cycle, and have low requirements for visual refinement—naturally suited for AI to cut into. But the same logic is moving upstream. Long-form dramas, web films, commercials, promotional videos—any category centered on "standardized visual output" is undergoing the same cost-accounting reshuffle.
Chinese Online's AI medium-length drama "The Demon Emperor's Butler" has already launched, and an AI animated drama based on the top long-form drama IP "The Legend of Shen Li" was also recently released; iReader Technology's AI medium-length drama in cooperation with a leading long video platform is expected to launch in the fourth quarter; Bona Film Group's AI theatrical film "Sanxingdui: Future Past" was already released in October, and a second AI film has also started production.
In a year of tightening budgets, no one will pay ten times more for the words "live-action shooting" unless the project itself is selling the actors' faces and the authenticity of the set. And those projects that can't sell either are disappearing from project lists.
This is also why this winter is different from previous ones. The film and television industry has experienced downturns before—tax rectification, platform price caps, pandemic shutdowns—and each round some people left, but each round some people persevered, because after the downturn demand would return, crews would restart, and those who left could come back. The problem with this round is that when demand returns, it may no longer be live-action crews that catch it.
After winter there is always spring, and the branches killed by frost will grow back next year. But AI's transformation of the film and television industry has no seasonal recovery. Those who left crews to drive ride-hailing cars, went back to their hometowns to farm, or changed careers to sales may never have the chance to return to a crew again.
For the 99 percent of ordinary people in the industry, they can only find another way out. Because this is not a winter that will pass, but a spring that will never return.