At noon on February 11, 2026, Li Tao, the founder of Xi’an Fengxing Culture, received a call from the docking person in charge of a short drama platform. The content was very short: All projects that have not been launched have been stopped.
Three years ago, “Wushuang” produced by Fengxing Culture exceeded 100 million yuan in recharge within 8 days of its launch, becoming a landmark event in the industry. By last year, the company’s production capacity of live-action short dramas had reached a maximum monthly output of over 100, with nearly 500 employees. The upcoming Spring Festival period had to be “filmed” as usual. After the call was made, the production line stopped where it was.
The system that shoots more than 100 movies a month has great inertia – sudden braking, how to stop, how to retreat, how to appease, everything is troublesome. Li Tao and the executives spent the entire holiday discussing how to deal with it after the new year.
Just after the Spring Festival, on March 3, information related to “many short drama projects have been suspended” became a hot search topic, and there were constant reports in the industry that short drama platforms were suspending live-action short drama projects in batches and canceling the producers’ guarantee commitments.
The so-called “guarantee” has been the foundation of this business in the past two years. The platform first “covers” the production fee according to the agreed amount. A short play usually costs 200,000 to 500,000 yuan, and the highest can be more than 1 million yuan. Generally speaking, the production company can earn 5% to 8% of the production profit, and the rest is left to the platform to face the market. If the drama “explodes”, the producer can also get a share of the project according to the contract. It is said that “less than 10” production companies across the country have guaranteed platform scale quotas, and all received notices around this node.
Just the day after this phone call, on February 12, ByteDance’s large video generation model Seedance 2.0 was officially released, and the price of video generation was slashed.
The direction of the story has changed. Statements such as “cutting costs”, “getting rich instantly”, “making a movie in three days and paying back the money in five days” and “everyone can make a drama” are circulated on the platform. With technological upgrades and platform policy adjustments, it seems that the short drama ecology will usher in a new wave of explosion.
Prior to this, the fastest person in the AI short drama Myth was Huang Haonan, the founder of Soy Sauce Culture. At its peak, eight of the top 10 dramas on Douyin were his dramas; at the end of last year, China Literature Group came to visit with IP and investment, and the company’s valuation was pushed to nearly 2 billion yuan. “This is a money printing machine.” Huang Haonan said that the revenue in 2025 will reach about 800 million yuan. At this time, he can still enjoy the last good period of his life, because two months later, he will personally lay off hundreds of people in the company with thousands of people.
“At that time, the platform went around telling others: Did you know that Soy Sauce made tens of millions by doing AI, and promoted it from house to house.” Huang Haonan said, “It treated people as ‘consumables.'” Set a benchmark and attract a large number of people to come in and do it. The result is that 80% of them will be “buried” and only 20% will make money.。
The industry had grown big enough before the brakes were applied. As of December 2025, the number of micro-short drama users is approaching 700 million, and the market size has increased from less than 4 billion yuan in 2021 to 50 billion yuan in 2024, and has exceeded 100 billion yuan by 2025.
But the other side is the high concentration on the supply side. QuestMobile data shows that in February 2026, the overall monthly active users of the short drama industry were 718 million, and the monthly active users of Hongguo Short Drama App exceeded 300 million. In July this year, Hongguo’s daily active users reached 168 million, a year-on-year increase of 107%, exceeding the combined daily active users of iQiyi, Tencent Video, Youku, and Mango TV.
According to people close to Douyin, from the end of 2025 to 2026, Hongguo will continue to move closer to Douyin’s parent ecosystem: the creation platform will merge with Douyin Group’s short drama copyright center, and content, contracts and data will be fully integrated. Companies on the chain will face a unified set of rules and a decision-making center. Every turn of the platform will be a “big visible hand” pushing people to different places.
In Xi’an and Zhengzhou, the cities with the highest production capacity, the number of crews working in each base on a single day has dropped from a dozen at the peak to one or two. In the first quarter of 2026, the number of live-action short drama starts fell by approximately 75% year-on-year.
Starting in March, companies that stopped filming began to collectively change tracks. Leading production companies in the real-shooting era, such as Xi’an Fengxing Culture, Zhuoyuan Film and Television, and Zhengzhou Sansheng Wanwu, have all shifted most of their production capacity to AI. In the first eight months of this year, 430,000 micro-short dramas were online online, 13 times the number last year, of which AI dramas accounted for more than 90%.
It stands to reason that as technology advances and costs drop, creativity and money-making will explode. However, according to DataEye data, based on the profit and loss benchmark of 50 million views, less than 1.3% of AI dramas were able to make back their money in the first half of the year, and only one out of every 77 new dramas was able to break even. Seven months have passed, and the most shouted cry in the industry is “not making money.” In the past six months, the most frequently asked question in the industry has changed from “how to do it faster and cheaper” to “should we do it again and what should we do?”
Why is it that the short drama has been replaced by AI instead of real people? Although it is obviously reducing costs and increasing efficiency, it is more difficult to make money?
Short dramas have been a highly layered industry chain since their birth: the upstream is web IP and scripts, the middle is the platform, producers, and distributors, and the downstream is the streamers and audiences. In the past three years, the power structure on this chain has experienced two rounds of drastic changes: First, the free model turns almost all links into platform suppliers, and then AI makes the “production” link itself lose its scarcity.
The deeper changes occurred in the ledger. The income of each link in the chain will ultimately be calculated by the platform, including which traffic pool a drama will enter, what coefficient it will be settled by, and how much a broadcast is worth. The rise and fall of the “settlement coefficient” affects everyone’s nerves.
Wang Xiaoquan, the co-founder of Zhoutu Culture, has experienced the most complete process of ups and downs in this round, and he still insists on focusing on live-action dramas. “If everyone doesn’t do real people, if I do it, it may be scarce.” He said: “The threshold for making an AI drama is very low, but the threshold for doing it well is very high.”
In the past seven months, the industry has verified the first half of the sentence, and the second half has just begun.
01
新規矩
It is understood that Seedance 2.0 has raised the generation price to about 0.6 yuan per second for 720P and about 1.5 yuan for 1080P. The production cost of AI comic dramas has been reduced to 800 to 1,200 yuan per minute, which is about one-fifth of that of live-action short dramas; the cost of an AI simulation drama has dropped to 60,000 to 100,000 yuan, while a live-action drama of the same size costs more than 500,000 yuan. If it is a “running drama”, the above costs can be reduced several times.
This kind of “cost reduction narrative” has been on the hot search for many times at the beginning of this year. Each company publishes a version of its workflow: shorter cycle, lower cost, fewer employees – 3 people, 5 days, one drama, which is regarded as the new industry standard; each iteration of a large model must be matched with a cheaper unit price. In just six months, the production quotations for standardized AI short dramas have dropped by more than 90%.
This directly detonated production capacity. In June, Douyin alone launched 37,700 new AI dramas, with an average of more than 1,200 per day. But the expansion of supply goes hand in hand with the dilution of returns: About every 77 new dramas, only 1 can make back its money. It is understood that the platform’s card approval rate reached 70% to 80% in the second quarter.
But this is all a story later. The industry quickly formed a consensus that AI was the next hot spot. After the Spring Festival, everyone is doing the same thing: changing tracks.
Li Tao puts his overall focus on AI. This was not easy, nearly 100 people left one after another; two partners also changed their jobs and became the general person in charge of the AI project. There was no ready-made tool, so the team simply wrote it themselves. In order to lock in the computing power, Fengxing signed an annual contract: the initial negotiation was for 10 million yuan, but when the contract was actually signed, the price increased to 15 million yuan. Several groups of agencies came to the door, all rushing to get him to sign. “There is no way, I can’t fall behind.” He still signed. Looking back now, there was also the possibility of “hungry marketing” routine at that time.
Some people turned around earlier. Liu Yihong, the founder of Xiduo Yuanyuan Film and Television, studied animation, and the company was formerly a film and television advertising company. As early as March last year, he had already led seven or eight people to test the waters of AI comics. The first work “Wanjie Qianzhuang” received more than 20 million views and paid off. He spread the team, and the AI team expanded to 50 or 60 people by July, and could produce 20 to 30 movies a month.
In July last year, Wang Hongfang, the founder of Zhengzhou Sansheng Wanwu, also selected a dozen people from the 100-person post-production team to try to use AI to make comics. There were no classes and no one to teach them, so they just tried the tools on the market one by one and polished the prompt words sentence by sentence.
In the first half of the year, there was “not much profit”, but the direction was set: no longer focusing on contract production, but turning to owning copyrights. He quickly came to his own conclusion about this wave: AI can only make money for people who really understand it.
As for what “understanding” means, Wang Hongfang believes that operations, resources, and technology are all indispensable. You can’t make money by just knowing technology, and resources alone are not enough. In the end, the test is operations, that is, issuance and monetization capabilities – “Where do you put it, can it turn into money.”
“Many people regard the capabilities of the model as the capabilities of their own company. This is a common problem in small companies.” He said that what is equalized in technology is tools, but what is not erased is cognition. “Only those with cognition can make money, and those without cognition are just cannon fodder.”
But the influx is growing. A local boss came to consult Wang Hongfang and wanted to get involved in AI short dramas. He thought it was an opportunity. He made a lot of money before, but now he wants to do something big – rent an office in a central location and recruit nearly a hundred people to do an AI skit. He poured cold water on the other party’s face, and his advice to the other party was always “small team, small scale”. He had already calculated the accounts for the other party: if he worked with 20 people, he would lose six months first, “that would be a loss of three to four hundred thousand. Let’s figure this out first.” But the other party didn’t listen, and later lost more than 2 million yuan in just 3 months.
Production organizations are being restructured. In the era of live-action dramas, the crew was a team of about 40 people, and the work was extremely detailed; the AI crew was reduced to less than 10 people, or even two or three people. The focus of the workflow is also shifting. The production of AI dramas is backwards from “post-production”: editors first figure out what scenes are needed for each segment, and then generate materials and make up for any missing parts.
In Wang Hongfang’s view, there are only two ways to survive this round: Either it is a leading company that “knows” very well, or it is a super individual that is extremely small – “With 3 to 5 people in a company, there is a way out.”. “Technical equality puts tools in everyone’s hands. “Ordinary people can make IP-based things at low cost and can polish it themselves.” However, companies that are caught in the middle and rely on layers of subcontracting are losing their position.
Others choose to stay on the live-action side. Speaking of which, Zhoutu’s reaction was actually very early: In the spring of 2025, Wang Xiaoquan collaborated with a game IP on a comic series, but it failed to be completed; in the summer, he increased the AI team to 70 or 80 people. But he has been unable to make up his mind whether to transform into AI as a whole. Zhoutu is an asset-heavy company, and the large amount of assets that have been suppressed during those years of live-action filming are carried like a baggage; a team of several hundred people cannot be dispersed easily.
At present, Zhoutu has shrunk to more than 100 people. It still produces 20 to 30 live-action dramas every month and 56 AI dramas. It has transformed from a short drama platform provider into a smaller and more flexible company. “Don’t do it blindly before you find the direction now,” he said. Live first.
02
Upstream changes
The script of the short play relies heavily on online texts.
Industry data agencies estimate that more than half of short dramas are adapted from Internet IPs, and the proportion of popular short dramas is as high as 80%. The license fee for short drama adaptations of S-level web IPs ranges from 80,000 to 100,000 yuan, popular A-level works range from 60,000 to 80,000 yuan, and the most common ones cost more than 30,000 yuan. China Literature, Tomato Novel, Chinese Online, and Migu Digital Media have all turned their IP libraries into “script pools” for short plays.
In the past two years, in order to lock good scripts into its own library, Hongguo has raised a round of prices for the screenwriter group: the minimum guarantee for the head screenwriter once reached 150,000 to 200,000 yuan, and the profit sharing is calculated separately. In contrast, other platforms have almost reached the top with a minimum guarantee of 120,000 yuan. The mechanism of the platform is to provide money from both ends: the screenwriter passes the script into the library, and only gets the money if it is selected; the producer has the authority to choose the script, and also receives the production fee.
This mechanism quickly created room for arbitrage: someone put the “relationship book” in the warehouse, and someone else picked it up, and the money started flowing after the contract was signed. Many practitioners mentioned that the platform has differentiated policies for different companies, and some large companies were almost “exempt from inspection”, but they were unwilling to name any companies. They only said that there will be too much money in 2025, and everyone is collecting it. “As soon as the platform settles the accounts at the end of the year, it will close the gap.”
AI makes the meaning of copyright even more complicated. Simulator dramas use a large number of faces that look like celebrities, and the approval rate has plummeted; the “AI script cleaning” that leaves the skeleton of a hit drama and re-makes it makes the marginal cost of imitating a hit drama close to zero.
▲Photography: Chen Hao
Upstream online writing platforms are also taking action in person. China Literature invested in Soy Sauce Culture in 2025 and also launched the “Comic Assistant” tool for comic production. Its 2025 annual report shows that more than 100 AI comic series launched in less than half a year have been viewed in excess of 10 million. Regarding this cooperation, Huang Haonan said very directly: “I also need the blessing of IP, and they also need a leading production company in the country.” IP holders are no longer satisfied with collecting licensing fees. They want to extend their hands into the next step of production, but the tooling of the entire industry is also reducing the scarcity of the “script adaptation” link.
There are also clear differences in the platform’s copyright attitudes towards the two types of content. According to people close to Douyin, the platform will tend to increase the proportion of exclusive and self-made copyrights in live-action short dramas to more than 60% in the next year or two. For AI short dramas, “it will only lock short-term exclusivity rights for innovative works, and most of them will only obtain distribution authorization, and will not touch permanent copyrights.”
The producer’s feelings are more “acute”. Huang Haonan said that practitioners in the entire industry are “Luotuo Xiangzi in the AI world”: “Not even Luotuo Xiangzi. Xiangzi has his own car, and we don’t even have our own car.He misses the days when platforms raised prices for each other: “In the era of Aiyu Tengmang, if a drama became a hit, I signed an exclusive contract with this one, and another one would come over immediately and buy another one at a higher price. What now? I had a blast at Hongguo, and other people could only watch. “Without competition in the market, copyright has no price.
The logic of the platform is not difficult to understand: the hit rate of AI dramas is only about 0.5%, with a short life cycle, strong replicability, and limited permanent copyright value; what the platform really wants is live-action dramas, which are scarce, can operate in the long term, and attract users.
This brings about a contradiction: the online literary platforms that control IP are appreciating in value, and the platforms that control traffic only reserve the promise of “permanent copyright” for scarce real-person content. The production companies caught in the middle, the more they use AI to improve efficiency, the closer they are to “replaceable production capacity” in the eyes of the platforms.
Wang Hongfang believes that the essence of the platform economy is that “when the number of users reaches 200-300 million, it will no longer need to purchase in large quantities; if it does not need to purchase in large quantities, it will not need to allocate so much money. This is a very normal business method.”
This also explains a seemingly abnormal phenomenon in the field of AI short dramas: platforms are not in a hurry to buy out the copyrights of AI dramas. The position of “copyright” in the AI era is being quietly replaced by “quotas” and “coefficients”.
03
標尺
Behind this “replacement” is that the platform has completed a contraction first.
In 2025, Douyin Group established a short drama copyright center to connect Douyin’s short drama section and Hongguo to the same channel. Previously, Hongguo was managed by Tomato Novels, while Douyin formed a separate line: one side has paid short dramas, and the other side has free short dramas plus advertising. After the establishment of the Copyright Center, everything from IP procurement to screenwriters, producers, and distributors has been integrated into one external interface.
According to people close to Douyin, the platform’s share for content owners is divided into “exclusive, first-release, sublicensed, and priority launch” categories. “Exclusive” can get the highest coefficient and long-term traffic. According to the official guidelines, transactions between platforms and content providers are divided into four categories: guaranteed production, full copyright for the entire library, full copyright for a single part, and non-exclusive. What the platform transfers the most is “guaranteed production” – the platform provides the script, pays the guaranteed fee and excess share, and obtains the intellectual property rights of the results. This model inspired a lot of production capacity last year.
In the second half of 2025, some companies even started production of more than 200 dramas in a month. If the crew was not enough, they would subcontract them at different levels; a drama worth 400,000 yuan can be filmed with 200,000 yuan. Two or three plays are filmed in one location, and the actors change their clothes and continue acting. What is saved is profit, but the quality of the film is overdrawn. This is also considered to be one of the incentives for the platform to cancel large-scale guarantees.
▲Photography: Chen Hao
According to the above-mentioned person, the content provider’s shared revenue is equal to “new viewing time × basic time unit price × content type coefficient × copyright coefficient × incentive coefficient × completion/rebroadcast correction coefficient”. AI dramas also have an additional “painting style coefficient”. Various “coefficients” are the starting point for the platform to regulate production. The “painting style coefficient” has been adjusted three times in the first half of 2026: in February, the coefficient for 2D comics was lowered, in April, the coefficient for human simulation and 3D comics was lowered, and in July, the coefficient for explanatory comics was raised. During the same period, the fusion drama of “live-action shooting + AI post-production” will receive an additional share incentive of up to 20%.
In the eyes of many production companies, this “dual-track system” is a benchmark: AI dramas require production capacity and new recruits, while live-action dramas require high-quality products and a basic base.. Starting from April 2026, Douyin has successively announced: 500 million yuan in special funds and a guaranteed budget of over 1.5 billion yuan; the share ratio of real-life paid content has been increased from 70% to 80%; under the “Thousand Movies Plan”, completed films that meet the standards can receive a guarantee of 200,000 to 4 million yuan.
The policy has also gradually become clear. On September 1, the “Measures for the Development and Management of Micro-Short Dramas” came into effect, and the “Moss Flower Label” was launched on the same day; on September 17, Wang Xiaoliang, director of the Online Audiovisual Program Management Department of the State Administration of Radio and Television, stated: “Real-life dramas are the main force in high-quality creation and must be vigorously supported.” He also required supporting measures on the platform – opening a special area for live-action dramas, real money support, and requiring labeling and authorization compliance for AI dramas.
There is another change in this round: the cooperation between platforms and live-action dramas is shifting from “distribution only” to going deep into the upstream of the industry. Provide money upfront, customize development, and participate in IP incubation. However, in many production companies, the physical sense has not kept up: the money that was originally available is no longer available, or is received less.
According to the above person, The core motivation for canceling large-scale guarantees and tightening account sharing is to make profits positive – the phase of subsidy exchange for scale is over, and this business will start to count profits.
According to Nomura Securities citing QuestMobile data, in July this year, Douyin’s total monthly usage time exceeded WeChat for the first time. However, according to Huxiu, ByteDance lowered its full-year advertising revenue and GMV expectations for its e-commerce business. Douyin later denied the relevant statement. According to calculations from public brokerage reports, Douyin’s ad loading rate is around 15%, approaching the upper limit, while WeChat’s video account is only between 3% and 5%.
In terms of short dramas, the GMV of Hongguo e-commerce in the first half of 2026 was approximately 15 billion yuan, and the payment GMV of Douyin e-commerce in the same period was approximately 2.4 trillion yuan. The same amount of traffic is allocated to short dramas in exchange for duration, not income.
According to common industry standards, the total pool of short drama platform advertising accounts is a formula: total exposure ÷ 1000 × eCPM (revenue per thousand exposures) × advertising fill rate. But the two key quantities in the formula are changing: exposure expands with supply and duration, advertising spaces increase dramatically, but unit prices are falling. The reporter learned from many practitioners that the eCPM of live-action dramas dropped from around 30 yuan in 2024 to less than 20 yuan in 2026, with the highest drop exceeding 40%.
The surge in AI production capacity has intensified dilution. The eCPM of AI dramas started at about 20 yuan. It once shot up and the unit revenue exceeded that of live-action dramas, and then fell back to about 14 yuan. According to widely circulated estimates in the industry, the ROI of AI dramas is generally only 1.03 to 1.07, while that of live-action dramas is 1.2 to 1.5. The more turbulent the supply, the cheaper the unit traffic. This may be the reason why the platform tightens AI accounting and improves the card approval rate.
In Huang Haonan’s view, the ceiling of this business is locked by two things: first, users’ time is limited, and second, the advertiser structure is single. “More than 90% of the advertisements in Hongguo are game advertisements. Will LV’s advertisements reach here? No.” High-end brands cannot enter, and the revenue from short dramas is tied to the gaming industry – the gaming market is only 300 billion yuan. “If the upstream is stuck at this scale and micro-short dramas can reach 300 billion, I cannot accept it logically.”
For producers, certainty is more important than direction. “The dividend period of wild growth has passed.” Wang Xiaoquan said, “But there should be a relatively certain expected income, and everyone can adjust production based on expectations.” The so-called “uncertainty” comes down to the specifics: whether there will be a minimum guarantee and whether the coefficient will be adjusted again. The calculation logic of ledger sharing is still a “black box” in their eyes. If the standards change every month, their optimal solution is to “stand still.”
There is also a saying among the producers that “the card reviewer is the card model”.
After staying at this card table for a long time, he is used to seeing the ups and downs. What Wang Xiaoquan is looking forward to now is a stable creative ecosystem.
He doesn’t even want other major companies to rush into the game: “When competition begins, there will definitely be no stability. Douyin will also provide defensive protection, whether it is supplier protection, IP protection or content protection.” He believes that only by establishing core barriers can healthy competition be achieved.
As for AI tools, he has a metaphor: “After the editing is released, the threshold for editing is lowered, and everyone can become a blogger.” However, he judged that the tools have smoothed the threshold, leaving a position that still belongs to professional institutions with high-quality mass production capabilities. This is what Zhoutu has been good at in the past, leaving the rest to the platform for now.
04
未定之局
Going overseas is another way out for practitioners.
According to DataEye data, in the first half of 2026, the top three overseas micro-short drama app in-app purchases are DramaBox, ReelShort, and NetShort, with approximately US$195 million, 188 million, and 128 million US dollars respectively. The top five are still all Chinese companies. But the situation is loosening: in the third quarter of 2025, among the TOP30 short drama applications, the download share of Chinese applications has dropped from 95% in the same period last year to 67%, and overseas local platforms have begun to fight back.
The financial reports of listed companies show that the profits of this track are far less impressive than the revenue figures. According to Chinese Online’s 2025 annual report, the company’s annual revenue was 1.657 billion yuan, and its net profit attributable to the parent company was a loss of 671 million yuan; Maple Leaf Interactive (the parent company of ReelShort), in which it is a shareholder, has high revenue, but is still losing money.
AI has added a new dimension to this track.
The first wave of dividends from going overseas is “translated dramas”, which translate and dub domestic live-action short dramas and send them overseas. AI makes the cost of this process close to zero: translation, dubbing, and lip-syncing are all handed over to the model.
80% of Fengxing’s AI dramas are exported overseas, but Li Tao bluntly said that they are still losing money – they do not have their own overseas platform and can only rely on other apps to monetize. He made his calculations very clear: Building an overseas platform by yourself is a money-burning game, and it would be difficult to do it without investing tens of millions of yuan a month.
Sansheng Wanwu also invests 80% of its production capacity overseas. At the end of March this year, Wang Hongfang decided to focus on overseas, with North America as the main battlefield. The company’s revenue in the first half of the year was about 100 million yuan. Although it was not as good as last year, he said that in the overseas drama track, they were still in the first echelon in China.
Zhengzhou’s Grain Harvest Culture mass-produces AI dramas for the European, American, Indian, and Mexican markets. The computing power and labor cost of a drama is about 20,000 yuan. Although the review cycle is based on weeks and the payment is slower, the income from 10,000 broadcasts overseas (the money distributed to the creators by the platform for every 10,000 effective plays) is several times that in China.
Nanchang’s soy sauce culture will start going overseas in April 2026, with overseas and domestic business accounting for half. It does not produce translated dramas, but customizes original content according to the tastes of European and American users. It first tests the waters on TikTok with themes such as vampires and magic schools that are rare in China.
But more people think that a self-made drama has been running overseas for a month, and the revenue is not worth a few days of computing power. Everyone on the chain finds his or her own place. Li Tao’s trajectory this year is almost a microcosm of the entire industry: It was the phone call at the beginning of the year, and he was still losing money in the summer. In the autumn, his judgment became simpler: AI has only changed the specific method of production. In the end, good content is still needed. “With good content, AI cannot replace people at all now.. ” he said.
Iteration of the model never stops. In June, Seedance 2.5 made its first appearance; at the end of July, it was officially released, extending the length of a single native video to 30 seconds; in August, the API was opened and a short drama workbench was already connected.
Related Reading
- Magic9 invites Alai, Honor will not be a substitute this time2026-09-30
- McDonald’s AI pricing push is widening the divide with some franchisees2026-09-30
- Yu Minhong doesn’t take Luo Yonghao as a stumbling block+2026-09-30
- Original Dong Mingzhu is “handing over” Gree2026-09-30
- Original Make a fortune in silence, Meitu earned 600 million in half a year2026-09-30
