Between Zhengzhou and Hengdian: The Cost and Survival Arithmetic of Live-Action Short Drama Factories Under the AI Short Drama Shock
In 2026, revenue sharing on live-action short drama platforms shrank sharply, and AI short dramas poured in at less than half the cost of live-action, forcing production companies in Hengdian and Zhengzhou to pivot to AI production. Chengfeng Polang Film's live-action short drama orders fell from more than 20 per month to around 10, and in April about 90% of its workforce shifted to AI production, with monthly output rising to 5,000 to 8,000 minutes from May to August. AI video generation is billed by the second at about 1 yuan; DataEye statistics show that in the first half of 2026, 221,900 new AI dramas and comic dramas were added on Douyin, and only 1.3% reached the break-even line of 50 million plays.
Topics
Body · AI translation
In mid-October, autumn was at its height in Hengdian, Zhejiang, and the whole town was filled with the fragrance of osmanthus. Walking to the end of Wansheng South Street brings you to the ticket gate of the film and television city. The electronic screen at the Qingming Shanghe Tu scenic area scrolled through the performance schedule; visitors were sparse, and some said it gets livelier at night. In the main hall of the Qinwang Palace, an actor dressed as a "eunuch" solicited visitors; tourists paid to sit on the dragon throne for a while and then took photos with him.
Leaving the scenic area, I went to a filming base in Hengdian that sells tickets to the public. At the entrance, the guard was gnawing on bread. Seeing me holding a ticket, he looked astonished: "People actually buy tickets?" He told me that no crews were filming inside anymore and there was nothing to see.
Across from a forest of commercial housing complexes, I found Deyu Film and Television. The staff member who received me said, "We just moved recently; this building has only been in use for a short time." Footsteps sounded on the stairs, and the company head, Tu Guanghao, pushed the door open and came upstairs. After sitting down, he remarked casually that another round of layoffs had happened in the past couple of days.
In 2025, Deyu Film and Television produced more than 400 live-action short dramas, while about 40,000 vertical-screen micro-short dramas were independently produced and released nationwide. Tu Guanghao said the company's revenue that year was about 160 million yuan. "Last year there was no need at all to worry about where the orders would come from—you just opened your mouth and the food was fed to you."
Just one month before I went to Hengdian, things had also gone quiet at the Dazhi Film and Television Base in Zhengzhou, Henan. The block used for Republican-era dramas was empty, prop vehicles sat in a corner of the yard gathering dust, and a cleaner sat on a plastic stool in the sun, saying the base was conducting fire safety inspections and everything had stopped. People at the base mentioned a building in Zhengzhou High-tech Zone that industry peers call the "AI Building." Most of the people inside are recent graduates, typing prompts at screens all day, repeatedly generating images, and stitching together video clips.
English text is being prepared.
The rough-and-tumble dividend of live-action short dramas has peaked, and costs are being calculated ever more thinly. Platforms are pushing budgets toward premium productions costing millions or even tens of millions, compliance is becoming more detailed, and shooting cycles are lengthening; small workshops accustomed to turning out work in a few days are struggling badly. On the other side, AI short dramas are squeezing in on the strength of computing power and models, probing entry through genres such as fantasy and science fiction that are extremely hard to realize in live action, with costs pressed to less than half those of live action.
Actors and behind-the-scenes personnel standing in the middle are the first to sense the change in water temperature. Since June, I have visited production teams and actors in Zhengzhou, Hengdian, Guiyang, and other places. Facing the explosion of AI short dramas, what they talk about most is always cost. Once it is clear how money flows among producers, traffic buyers, and platforms, the industry's pattern of interests and the rise and fall of the people in it also become clear.
Platforms Slam on the Brakes
In August, Chengfeng Polang Film was working in an office building in Zhengzhou. When the conversation grew heated, partner Zhang Pengfei sat across the table, taking sip after sip of tea and stubbing out cigarette after cigarette.
2025 was the year costs rose fastest. Zhang Pengfei said top-tier actors earned 50,000 yuan a day, second-tier actors 3,000 or 4,000 yuan, and actors in the second or third roles 1,500 to 3,000 yuan a day. A drama shot over seven or eight days saw costs rise from 200,000 yuan in the early period to 1.2 million or 1.5 million yuan, and quotations of two or three million yuan had appeared on the market. For the same ordinary actor, a quote this week might be 8,000 yuan; next week, after appearing in a hit drama, asking again would bring 16,000 yuan. Quotes for lighting equipment, costumes, makeup, props, and extras also rose accordingly.
Production companies dared to invest because of the platforms' purchasing and revenue-sharing arrangements at the time. Zhang Pengfei explained that some projects had production fees or guaranteed minimum arrangements first, and after the work went online, revenue was shared according to performance. From 2023 to 2024, platforms were eager to expand the volume of releases; some platforms put 200 to 500 dramas online a month, relying on quantity to find hits. "As long as one or two dramas break out, all the costs can be covered."
Zhang Pengfei called the months from December 2025 to March 2026 "the industry slamming on the brakes." Some live-action short-drama projects undertaken by Chengfeng Polang lost their original guaranteed minimum arrangements, and the revenue-sharing method also changed.
Around February 15 was when Hongguo issued its statements. From 2 a.m. on the day the statement came out, Zhang Pengfei's phone rang nonstop, with partners demanding that "projects without signed contracts be stopped now." He said the revenue share on the statement was roughly half of what it had been. A drama that generated 50 million in popularity could previously share 1 million to 1.5 million yuan; at that time it became 400,000 to 600,000 yuan.
"The New Year season in previous years was actually this circle's carnival night. After all, at that time everyone had leisure and money and was willing to spend. That period was when every drama recovered the most, but last year's New Year season was simply a disaster," Zhang Pengfei said. One live-action short drama Chengfeng Polang made for the New Year season cost more than 1.4 million yuan and reached third place in popularity across the entire internet, yet first-month revenue sharing did not recover the cost.
Projects with contracts signed earlier still shot as usual in March, and the impact of the adjustment only became concentrated in April. Zhang Pengfei said the company's live-action short-drama orders fell from more than 20 a month to about 10. The number of crews at some filming bases in Zhengzhou also dropped noticeably. In the spring of 2026, Chengfeng Polang shot a costume short drama in Hengdian, and almost no other crews could be seen in the scenic area where it was located.
Zhang Pengfei had previously heard that platforms might cut guaranteed minimums and had expected the adjustment to proceed gradually; the actual scale and speed exceeded his expectations. He believes platforms need to control the relationship between content investment and recovery, but still need quality works. Tu Guanghao also believes long-term partner production companies may still receive orders.
QuestMobile data shows that in August 2026, Hongguo Short Drama reached 403 million monthly active users and 177 million daily active users. The user base remains huge, yet production companies are feeling the contraction in live-action short-drama purchasing.
Chenfeng Polang's attempt at AI short dramas began in November 2025. At the time, orders for live-action short dramas were still plentiful, and founder Liu Tengyuan had already proposed developing an AI business. In December, an AI team of two or three people was formed; the first simulated live-action short drama took about a month to make, cost nearly 50,000 yuan in computing power, and failed to turn a profit. The second was slightly profitable, and the third's production cycle was shortened to 20 days, with revenue reaching four times the cost.
After the 2026 Spring Festival, the experience from the early experiments paid off. In March, the company formed a preliminary AI short-drama production workflow; in early April, about 90 percent of its workforce shifted to AI production. Zhang Pengfei said that by April the company could produce more than 20 AI short dramas, and from May to August monthly output rose to 5,000 to 8,000 minutes.
Quotes for live-action short-drama projects also gradually split toward the low-budget and high-investment ends, with mid-range projects decreasing. The "Thousand-Project Plan" project undertaken by Deyu Film and Television was quoted at 180,000 yuan per title. Tu Guanghao did the math: shooting for two days still left a small profit, while shooting for three days roughly broke even.
A new production system has taken shape
In an open-plan office area, young employees are typing at computer screens, with several large-model interfaces, Feishu multi-dimensional tables and CapCut open on the screens. Bai Jingyi led me forward along the workstations, from script production and digital asset generation all the way to the editing area at the back. Production is roughly divided into three stages: script processing, image generation and post-production editing.
Scripts come from the in-house screenwriting department and also from partner screenwriting companies. The director first reads through the script, then hands it to AI to break down into a storyboard script, and then checks shot continuity, character actions and camera language one by one. Bai Jingyi said AI can break down a script, but how the story advances still needs the director's oversight: "Whether it's AI or variety shows, the director is the core."
Once the storyboard is set, the system continues to compile the digital asset list. How many characters there are, in which scenes each appears, what clothes they wear, and which scenes, props and special effects are needed must all be listed in advance. In the Feishu multi-dimensional table, the protagonist's appearance, hairstyle and clothing are categorized as character assets; scenes such as the main hall, streets and dungeons belong to scene assets; objects such as swords and hairpins belong to prop assets. Production staff, reviewers and the director share the same table to check generated materials.
Dafei, who is in charge of AI short-drama production, said production staff first need to understand the character's personality and the story setting, and then translate the text description into images. The company uses tools such as the GPT image model and Banana to create static images, and stores them in the asset library once satisfied. Scenes and some supporting characters can be reused, while protagonists usually need to be redesigned for each new script. The more accurate the assets, the easier it is to subsequently generate images that meet requirements.
Next, production staff combine character and scene images with the storyboard script, write in prompts for character actions, camera movement, lighting and emotion, and hand them to the video model to generate dynamic footage. The company has accumulated prompts for different genres, and each team also keeps adjusting them.
Production staff call video generation "drawing cards." The model generates clips of a few seconds to a few dozen seconds at a time, and staff check them segment by segment. When a video clip fails to meet requirements, the card-drawing operator revises the prompt and regenerates it. Static images can still be adjusted in detail with Photoshop, but video involves continuous action and emotional changes, and fight scenes and complex actions are especially hard to control.
The generated video clips are screened and then handed to post-production editing. Bai Jingyi explained that a production line usually has five or six people: the director controls the storyboard and final result, while others collaborate on asset production, video generation, review and editing. The company also produces 3D CG, simulated live-action and other types, and assigns projects according to the genres each team excels in.
Dafei summed up the whole method as three questions: who, where, and what happened. Character assets answer "who," scene assets answer "where," and video generation is responsible for showing "what happened." But how much footage can be generated can be counted by model call volume; whether a story can attract viewers still requires human judgment, and ultimately depends on how it performs after release.
On Chenfeng Polang's AI short-drama team, many staff are junior-college graduates who have just finished school. They are familiar with model operation and editing tools and can carry out generation tasks according to their division of labor. Traditional film sets rely on experience in cinematography, lighting and on-set scheduling, whereas AI production shifts part of the work to the computer. People familiar with cinematography techniques may not quickly adapt to this pace.
At Dazhi Film and Television Base, logistics staff told me that the easiest people to transition are editors who originally did post-production, because they can sit still in front of a computer and know exactly what footage a shot needs. The base once reassigned a camera crew of 20 to 30 people to an office to test AI short-drama production, and before long they had all been sent back to the live-action equipment warehouse. The camera operators complained that on set they could already have shot two or three episodes, while generating a single shot in front of a computer takes 5 to 10 minutes of waiting, and if you miss writing in an eye light you have to wait all over again—by the end of a day, they might not finish even one episode.
In September, I asked Wu Hongbing, head of Dazhi Film and Television Base, whether I could visit their AI short-drama team. He replied: "It's basically disbanded now; we can't keep it going."
Traditional film and television production relies on master-apprentice training for technical staff. Staff at Dazhi Film and Television Base said the company requires camera assistants to work for at least three years, during which they learn cinematography and lighting techniques and can operate a camera independently only after mastering them. After short dramas took off, some people without professional training also entered film sets. Shan Ge, who has been a lighting technician for 10 years, is very unhappy about this and feels the industry's original rules have been disrupted.
Shan Ge has been doing lighting for 10 years. Seven or eight years ago, he was in charge of lighting the entire street exterior for an online video project, leading as many as 12 people; nowadays he usually leads only 3. He says the crew's budget for the lighting department is only enough to arrange that many people. On September 17, he had already worked 15 days that month and still had to go on set in the afternoon. At most he can work 28 days a month, and some peers cannot even get 8 days of work in a month. Lighting technicians are paid by the number of days worked; with no work, there is no corresponding labor income.
Dazhi Film and Television Base still has orders. Staff said the base receives an average of two or three crews per day, fewer than last year, but advertising, promotional videos and TV program shoots have never stopped. The Henan TV "Chinese Festivals: Wonderful Tour" series, State Grid promotional videos, and promotional videos for universities such as Henan University and Zhengzhou University are among the businesses they mentioned. Short-drama shoots commonly use the Sony FX3, while advertising and promotional videos more often use equipment such as ARRI Mini and ARRI 4D. The base still needs to rely on cinematography, lighting teams and equipment to take on these traditional live-action projects.
The LED lights in the warehouse were purchased in batches at a cost of several million yuan after short dramas rose to prominence. After orders declined, the equipment still needs maintenance and depreciation. For Dazhi Film and Television Base, whether the equipment bought during the short-drama peak can continue to generate income depends on whether it can obtain enough orders from advertising, promotional videos and other live-action projects.
The money flowed to computing power
Over the four months from May to August, a short-drama production company spent nearly 2 million yuan on AI video generation, all settled with Volcano Engine, with generated footage uniformly at 720P. The company's head said the computing cost for a drama was 25,000 to 30,000 yuan in the early period and later declined somewhat. At the service prices used at the time, generating 1 second of video cost about 1 yuan, and a single 15-second generation cost 15 yuan.
Computing power has become a separately calculable cost in AI short-drama production. Volcano Engine's publicly disclosed billing instructions show that some video generation models charge by generation duration, with prices for different versions ranging from a few tenths of a yuan to 1 yuan per second. For production companies, how much footage is generated and how many times it must be redone both affect final spending.
Operations by different personnel bring obvious cost differences. People from an editing background know what shot sizes are missing from the footage and make only targeted additions, consuming less computing power; inexperienced people easily redo entire segments. Production cycles are also shortening, from 20 to 30 days early on to half a month, a week, and as fast as 5 days to deliver a film. Higher production efficiency means more series can be produced in the same amount of time, and the industry's content supply increases accordingly.
A short-drama production company in Guiyang is also looking for ways to reduce computing expenses. The company changed computing channels three times in just over two months and finally obtained an 80% discount interface through an agent in Hebei, with the underlying model still connected to Volcano Engine. Computing costs also depend on production precision: rough production costs close to 200 yuan per minute, while pursuing higher image quality can push it up to over a thousand yuan. Even after lowering production costs, this company still lacks a mature traffic-buying partner, and whether its works can recoup their costs remains unknown.
According to DataEye statistics, in the first half of 2026, the Douyin platform alone added 221,900 AI dramas and animated dramas, of which 1,055 surpassed 100 million views, accounting for 0.48%. Under the break-even baseline of 50 million views set by that institution, 2,886 works reached the threshold, accounting for 1.3%. This statistic covers only the Douyin platform and cannot directly represent the actual profitability of the entire industry, but it can reflect the situation in which a large number of new dramas find it difficult to obtain enough views on a single platform.
After a film goes online, it still has to pass the traffic-buying stage. The content department and the traffic-buying department of Chengfeng Polang Film Industry are assessed separately; the traffic-buying department looks at return on investment and will not prioritize placement just because a drama is produced in-house—the key is whether the money can turn positive. Tu Guanghao believes that with a large number of dramas launching at the same time and without paid traffic support, it is very difficult for works to break out through organic traffic. The founder of a short-drama company said that producing for others brings a fixed fee and no traffic-buying risk; producing with one's own money, however, often makes it hard to recoup costs after entering the bidding for ad purchases.
Production companies can increase capacity by adding staff and buying computing power, but they cannot guarantee that every drama they produce will obtain corresponding traffic and revenue. Production costs occur before a work goes online, traffic-buying costs occur after it goes online, and final returns depend on whether viewers are willing to watch and how the platform distributes revenue.
Bai Jingyi stood beside her workstation looking at the screen. She said short dramas have a strong self-media character, and the topic selection must first be able to attract viewers. Technology has lowered the production threshold, making story and topic selection even more important. According to DataEye statistics, among the top 30 AI dramas and animated dramas by Hongguo popularity in the first half of 2026, 19 are IP adaptations. Mature story resources account for a relatively high proportion of top works, which has also led some production companies to reconsider the value of copyright and content development.
Changes in production costs have also affected the bargaining relationship between platforms and production companies. Zhang Pengfei said computing power prices can be checked, and staff wages and production days are also easy to estimate. With 30,000 yuan for computing power and 30,000 yuan for labor, totaling 60,000 yuan, the platform will ask whether it can be done for 55,000 yuan. Production companies also have to bear wages, rent and taxes during gaps between orders; costs do not drop to zero immediately when a project ends.
Companies such as Jiuzhou, Zhangyue, Dianzhong and Maiya have set up screenwriting or production operations in Zhengzhou, and some platforms and content companies have increased the share of self-produced short dramas in their business. Zhang Pengfei believes the cost advantage of pure contract production is shrinking, and smaller teams in particular find it difficult to obtain stable orders. He hopes to gradually shift to producing with his own copyrights and then distribute through external platforms.
A production company's operations are also affected by adjustments to platform rules. Zhang Pengfei remembers that the platform once indicated that dramas within 120 minutes had better completion rates, so the team cut finished films to under 120 minutes. Later, new rules set different revenue-sharing coefficients for different durations: dramas over 120 minutes were calculated at twice the coefficient, and those under 120 minutes at 1.2 times. On July 28 they received notice requiring completion by the 31st, and the team worked around the clock for three days, re-editing, adding duration and uploading the dramas.
Internet companies change so quickly that some in the short-drama industry joke: "One cut a month, and no two cuts are alike."
Cost determines location
On June 27, at an industry exchange meeting at the Sheraton Zhengzhou, an employee of a film and television branch talked about shooting costs in Xi'an. Its parent company's main business is steel, and its short-drama business is based in Xi'an. Local shooting-location fees had risen from 800 yuan to 4,200 or 4,300 yuan per day. The team did the math and decided to send the director, cinematographer, actors and production coordinator to Hengdian together, rent housing for two months and shoot four dramas consecutively; the overall cost was actually lower.
Hengdian has ready-made replica ancient streets, prop warehouses and soundstages, and it is easy to find actors, lighting technicians, cinematographers, costume, makeup and prop staff. Crews can shoot continuously, saving the time needed to find locations, rent props and coordinate personnel on short notice. When choosing where to shoot, production companies compare not only rent and wages but also how many people and how much equipment a drama requires, and whether those resources can be found in time.
Zhengzhou was able to develop a short-drama industry because it had its own foundations. Around 2018, Zhengzhou and Xi'an had already gathered a group of teams producing feed ads. They adapted exciting plots from web novels into short videos and used ad placements to attract users to read the original works, accumulating experience in shooting, editing and traffic operations.
Zhengzhou also had an important industry backdrop. During the real estate boom, large-scale property development and sales created demand for services such as advertising planning, video shooting and event execution. In 2018, commercial housing sales area in Zhengzhou reached 37.12 million square meters, then gradually declined to 19.09 million square meters by 2024, nearly halving. As the real estate market continued to adjust, related marketing service industries also faced pressure from shrinking client budgets.
Several industry insiders said short dramas happened to rise during this period, and some teams originally in advertising, wedding services, MCN and other businesses gradually transformed, bringing personnel, equipment and shooting experience into the new industry. The formation of Zhengzhou's short-drama industry rested both on the technical and talent foundation accumulated from feed ads and on a group of content producers seeking new orders after demand in local traditional businesses changed.
According to local public figures, by the end of 2024 Zhengzhou had more than 820 short-drama production companies, and by 2025 the number exceeded 1,000. As the number of companies grew rapidly, even choosing a registered company name became increasingly difficult. In earlier years, two-character company names were fairly common; later, three-character names became harder to register, and some newly established companies could only consider four-character names. The change in company names reflects, from one angle, the entrepreneurial heat of Zhengzhou's short-drama industry.
After companies clustered together, actors, cinematographers, screenwriters and post-production staff could move between projects, and production companies found it easier to find partner teams. Zhengzhou thus formed a network of personnel and services around short-drama production. However, compared with a traditional film and television base like Hengdian, Zhengzhou still lacks sufficient professional film and television equipment.
Staff at the Dazhi Film and Television Base said that Zhengzhou's stock of professional equipment is smaller than that of Beijing, Shanghai and Hengdian, and equipment rental frequency is also lower. The same camera might be rented frequently at a mature studio lot, but in Zhengzhou its schedule may not be fully booked. After equipment is purchased, costs must be recovered through rental; if utilization is low and depreciation is spread across each working day, the rent may actually be more expensive.
With the rise of AI short dramas, production companies have new options. AI short-drama director Li Yunming previously worked for an employer in Beijing, and the investor later arranged for him to return to his hometown of Jilin to set up a branch to reduce operating expenses. Tu Guanghao, meanwhile, set up AI production teams in Shenyang and Zhengzhou outside Hengdian, also attracted by relatively low local wages and labor costs.
Live-action short dramas require actors, soundstages, lighting, costumes and props, and much of the work must be done on set. AI production moves part of the work in front of computers, and production staff can call models and generate images from different cities. Companies can therefore arrange teams according to local wages and office costs without concentrating all personnel in traditional film and television bases.
The experience of a company in Guiyang offers another sample. The company has more than ten years of film and television production experience and once invested hundreds of thousands of yuan in shooting live-action short dramas, with actor costs approaching 200,000 yuan. The team had to look for locations in Huaxi and Guanshanhu, and props also had to be prepared on short notice. In June, staff told me that few teams producing live-action short dramas could still be seen in Guizhou. Guanghe later shifted to AI comic dramas and developed an internal production workflow platform to reduce reliance on live-action locations and actors.
In 2025, Hengdian launched 4,016 new vertical-screen micro-short-drama crews for the full year. Jiefang Daily reported that as of the end of May 2026, Hengdian had hosted 1,499 short-drama crews, down at least 50% year on year. Entertainment Capital, based on crew updates published by Hengdian, calculated that the number of vertical-screen short-drama crews in production in the third quarter fell 82.5% year on year. In the same period last year, as many as 106 crews were shooting in a single week; this year, the maximum in the same period was 24. The two sets of data use different statistical scopes, but both show a contraction in Hengdian's live-action short-drama shooting business.
In June, actor Wu Weibin mentioned that a group of Xi'an actors had moved to Hengdian in search of work opportunities. By mid-October, when I visited Hengdian, it was already hard to see short-drama crews shooting; only a few scattered teams were rushing to finish work. Before going to Hengdian, I asked Ren Lei to recommend a company that had once shot short dramas locally. After getting in touch, I learned that the company had already left Hengdian.
Hengdian is also looking for new uses for its studio resources. Experience programs such as "Into the Scene at Hengdian" let tourists choose a script, put on costumes, and complete a shoot with the help of directors, cinematographers, and makeup and styling staff. During the 2026 National Day holiday, an experience ticket including admission, shooting, and makeup and styling cost 549 yuan. Sets, equipment, and personnel that originally served drama crews have begun providing shooting services for tourists.
In October, a consortium formed by Dongyang Deep Space Engine and Volcano Engine won the bid for the Hengdian Film and Television AI Service Platform construction project at about 230 million yuan. On October 23, Hengdian launched the "Hengdian π" film and television AI service platform, which plans to integrate film and television large models, digital assets, creation tools, and computing resources.
The crews dispersed, and computing power moved in. The business on the studio lot has taken on a different look.
Retaining cash flow
On October 15, Tu Guanghao had just finished a meeting with finance. Deyu Film and Television was going to lay off staff again.
In 2025, the company had more than 200 employees at its peak, with annual revenue of about 160 million yuan. In 2026 it carried out four rounds of layoffs, and by the time we met, only about 30 people remained. Tu Guanghao was reluctant to let go of the professional production staff he had trained for a long time, but the financial statements showed that the company could no longer afford the previous level of spending.
Deyu's original office was more than 1,000 square meters, with annual rent of more than 260,000 yuan and monthly utility bills of nearly 20,000 yuan, totaling nearly 500,000 yuan a year. Now it has moved into a residential building, and the annual rent has dropped to 55,000 yuan. Tu Guanghao spent 300,000 yuan of his own money on renovations, and utilities are also paid at residential rates. After the staff reduction, the company still needs to pay more than 200,000 yuan a month, with social insurance calculated separately.
He plans to keep key positions such as business, directing, production, and post-production, as well as the necessary administrative and finance staff. The company still has several partners, each responsible for different business areas. Tu Guanghao said that if things get really tough, keeping a few partners plus administration and finance would be enough to keep the company going. Once orders increase, they can hire again.
In October, Deyu planned to shoot six live-action short dramas. Tu Guanghao estimated that the company spent about 600,000 yuan last month, and even with revenue sharing from previous works, that may not be enough to cover expenses. He hopes that after further staff reductions, the company can maintain basic operations on the orders it has on hand.
Three years ago, when Tu Guanghao started his business, he was much bolder about spending money.
He came to Hengdian in 2009 and worked as an actor agent and crew coordinator. Around 2019, as the traditional film and television industry was in a downturn, he began trying information-flow ads and novel promotion videos. The first time he shot a short drama for Dianzhong, the platform put up 120,000 yuan. He made no profit on the production cost, but in the first month after release he received 200,000 to 300,000 yuan in revenue sharing. He felt this business was viable.
Later he founded Deyu Film and Television. For one drama, the platform put up 100,000 yuan, and he invested another 180,000 yuan himself. To raise the money, he sold a Mercedes-Benz E300 for 240,000 yuan. He had bought that car in 2018 for more than 480,000 yuan. He also rented an office of more than 300 square meters, bought computers, and built a team.
In its early days, the company took production orders of 150,000 yuan and sometimes added another 100,000 yuan itself. Tu Guanghao wanted to make the works well and earn more income through revenue sharing after release. Deyu later produced hit short dramas and consequently received more orders. By 2025, the company could produce more than 30 live-action short dramas a month.
In 2026, he tried investing in and producing short dramas himself, putting in several million yuan in total and recovering less than half. One work cost less than 400,000 yuan to produce and returned about 15 times that, but the money earned could not cover the losses on other projects. Tu Guanghao suspended the remaining self-produced plans and continued taking platform custom projects.
A white paper released by the China Netcasting Services Association in November 2025 showed that there were 100,200 existing enterprises related to micro-short dramas in China, with 16,800 newly registered in the first nine months of 2025. The number of companies keeps growing, but their operating conditions vary. Deyu once earned income by expanding production scale, but after orders declined, its large team instead continued to consume funds.
Tu Guanghao is still willing to take on some projects with very thin profits. The platform pays 180,000 yuan in production fees; shooting for two days still leaves a small profit, while shooting for three days only roughly breaks even. He said the company has staff who have worked with it for years and has also signed some actors, so he cannot let everyone be idle indefinitely.
There is also a time lag in project payments. Before the 2026 Spring Festival, some production fees from the platform had not yet been settled, so Tu Guanghao advanced the money and paid the staff in full. I asked him why he did not wait for the platform to pay. He said: "If the platform doesn't pay me, that's between me and the platform. But they come to work for me, and that's between me and them."
Deyu can keep operating thanks to another source of income from works already released. Tu Guanghao said some hit dramas produced in 2025 are still generating continuous revenue sharing on Hongguo, with money arriving every month. He hopes to use this income to offset daily expenses such as salaries and rent, so the company can temporarily maintain a balance between income and expenditure.
This kind of income comes from works accumulated earlier, and it also requires those works to keep generating returns. For production companies without hit dramas and lacking financial reserves, waiting for orders to recover may be even harder.
Tu Guanghao remembers the trough in the film and television industry around 2019, when some directors went back to running street stalls, and action choreographers and costume and makeup staff switched to selling houses or delivering food. After short dramas took off, many people returned to the sets in Hengdian. He hopes Deyu can also wait for the rise of the next cycle.
Someone asked him about his company's three-year and five-year plans, and he found it hard to answer. The short drama industry changes too fast, and being able to arrange operations by the quarter is already quite good. What he can do is lower monthly expenses, keep the people who can still take on projects, and preserve the company's cash as much as possible.
"The boss still needs money in his pocket. If you have no money, when the industry gets better, what will you use to bring the company back to life?" The goal Tu Guanghao set for 2026 is simple: "Not losing money this year is making money."
Goodbye, Hengdian
Wu Weibin has a habit of keeping accounts. Every time he acts in a drama, he records the title, role, pay, referrer, and whether the money has been settled. From 2023 to June 2026, he estimates he acted in more than 170 short dramas. Looking through these records, the changes in the short drama industry ultimately come down to how much he can earn each month.
In 2023, Wu Weibin was still in Shanghai. At a film and television industry networking event, he heard that someone invested 100,000 yuan to make a short drama and ended up making 7 million yuan, so he wanted to try his luck in Hengdian. He found a notice in an actors' group chat, bought his own high-speed rail ticket, went there, and shot for three days at 180 yuan per day. The production team did not reimburse travel costs, so after calculating everything he actually lost money.
But he confirmed one thing: there were short dramas to shoot in Hengdian, and actors could get paid. At the end of June that year, he went to Hengdian again and gradually stayed on.
There were more and more short drama notices, and Wu Weibin became busier and busier. By the end of 2023, his daily pay had reached 300 to 400 yuan. In 2025 it rose to 1,500 yuan, and some roles could pay 2,000 yuan. In May and June, he went 61 consecutive days without a break, and in ordinary months he could still work more than 20 days.
Short dramas have short shooting cycles, so actors can take jobs across different production teams. Wu Weibin remembers that one day he rushed to three production teams, playing one role in the morning, switching to another in the afternoon, and not finishing until after four in the morning the next day. Sometimes a scene took only half an hour to shoot, and he could still get the agreed daily pay.
In early April 2025, his schedule was already full through early May, and he even posted on WeChat Moments to tell fellow actors that he was fully booked that month and to contact him the next month.
During that period, producers and casting staff would take the initiative to contact him. He did not need to go to production teams every day to submit materials and audition, nor did he need to search repeatedly for notices in actors' group chats. Wu Weibin said he was certainly happy about the higher income, but what satisfied him more was that he finally no longer had to be picked over by others.
After the 2026 Spring Festival, fewer people contacted him to act.
In March, he worked 5 days; in April, also 5 days; in May, only four or five days. Daily pay for some projects dropped to around 800 yuan. In his lowest-earning month, he made only 4,000 yuan.
Wu Weibin began actively looking for work again. He contacted directors and casting staff he knew and also signed up in actors' group chats. Some projects that had been said to start shooting were later canceled or postponed. Occasionally a production team was recruiting, but many actors signed up.
The drop in daily pay had already reduced his earnings, and the impact of fewer working days was even greater. Even if it were still calculated at 1,500 yuan a day, shooting only 5 days a month would still amount to only 7,500 yuan. Actors' income depends on actual days worked, and time spent waiting for notices is unpaid.
Living costs in Hengdian still had to be paid as usual. Wu Weibin calculated that rent plus utilities came to about 1,000 yuan a month, and daily expenses such as food required another 2,000 yuan. Earning 4,000 yuan a month was barely enough for himself, and it was hard to send money home.
He has two children still in school. When income was good, every time he received his pay he would transfer money home and keep only a small amount for living expenses. Now he sometimes has to borrow money to cover rent.
When he had no work, he often went to an internet cafe. At first he went to Sihe Internet Cafe, and later switched to Baxian Internet Cafe because the latter cost only 1.5 yuan per hour on average. He calculated that it was even cheaper than staying in his rented room with the air conditioner on.
From a daily fee of 1,500 yuan to calculating the hourly cost at an internet cafe, Wu Weibin rearranged the rhythm of his life.
He had shot more than a hundred short dramas in Hengdian and knew many directors, producers, and actors. Those connections had once helped him obtain stable work opportunities. After production teams decreased, acquaintances also found it hard to introduce him to new dramas.
In June, Wu Weibin and I chatted online once. At the time he was shooting a short drama in Qingdao, Shandong. That month he took part in several projects in succession, and his income recovered somewhat compared with before, but he rarely stayed in Hengdian anymore.
Short drama production companies can control expenses by laying off staff and reducing investment in locations and equipment, but actors have limited room to adjust.
On October 4, he posted on WeChat Moments: "For a long time to come, I will no longer act in vertical-screen short dramas. Please be informed."
On October 14, I arrived in Hengdian and wanted to invite him out to meet and ask what exactly had happened.
He told me he had already left Hengdian.
In September, Wu Weibin returned to his hometown in Shandong and currently works as an NPC and theater actor at the Shuihu Liangshan Scenic Area. I asked him why he chose to leave Hengdian.
"Fallen leaves return to their roots," he said. "After so many years of drifting, I'm tired. Four years drifting in Beijing, ten years drifting in Shanghai, three years drifting in Hengdian. Next month, in November, I'll be forty years old. I don't have much energy left to keep struggling, so I just want to be close to home and spend more time with my children and family and parents."
This article comes from the WeChat public account "Xiutai," author: Shi Can, published with authorization from 36Kr.