Artificial intelligence enters the era of “prompt words”

“We just want to quietly do drama, but now the frequency of interviews is a bit high.”

This is a sentence Li Chaohua, the chief producer and chief screenwriter of “Armed into the Song Dynasty” said repeatedly during a recent interview. A sense of fatigue was hidden behind the words. This small grassroots team, which originally only wanted to quietly produce content, was pushed to the forefront by an unexpected wave of traffic.

This drama, which is said to have no paid streaming throughout, has become an extremely rare example in the industry.

On August 21, the AI ​​comic short drama “Armed into the Song Dynasty” was launched on the Hongguo short drama platform. It has been online for more than two months and has been continuously updated for ten seasons and nearly 1,000 episodes; the platform collection has exceeded 11 million, and the cumulative broadcasts on the entire network have exceeded 12 billion as of November 7. The highest popularity in a single season is 82.06 million, and nearly 4 million people have reserved follow-up episodes.

In the context of the AI ​​short drama industry, this achievement is particularly eye-catching. Industry data shows that in the first half of 2026, there were more than 221,900 new AI short dramas launched on Douyin alone, and only 0.47% of dramas with over 100 million views were played. The industry generally regards 50 million views as the break-even line, and less than 1.3% of works can cross this line.

Recently, the Chinese pinyin “Duanju” has become popular overseas and has been cited by many overseas media to specifically refer to the micro-short drama category originating from China. This means that Chinese micro-short dramas, as the native audio-visual form of the Internet, are being recognized by the global industry.

穿越故事

Time travel stories have been tried and true, and the reason behind them is still human nature, that is, to escape the shackles of reality and enjoy grudges and grudges.

The same goes for “Armed into the Song Dynasty”, but it respects history more. This short play is adapted from an online article of the same name on the Tomato platform. The story is anchored in the fourth year of Xuanhe in the Northern Song Dynasty, only five years before the Jingkang Incident. A modern five-man special operations team accidentally travels to Qinghe Village in the Song Dynasty. There is no system panel or unlimited resources and funds that are common in Internet literature. There are clear limits on the ammunition, fuel, and medicines that can be carried with them. The members of the team relied on the modern division of labor system, tactical thinking, engineering and medical knowledge to farm fields, train troops, and build defense lines on the spot. They were passively involved in the vortex of the game between Song, Xia, and Song and Jin, trying to find the possibility of breaking the situation within the established historical trend.

Different from the narrative logic of most time-travel dramas on the market that crush the ancients with one click, this drama has a strong sense of restraint: bullets will run out, supplies will be difficult to sustain, and it is difficult for individual power to reverse the long-standing shortcomings of a huge dynasty. The creative team put a lot of effort into historical details. The quality of official uniforms, the household registration system, court etiquette, military and political structure, and grain and grass logistics rules were all based on Song historical materials. The characters and scene assets in the play were repeatedly polished to avoid the common problems of confusing faces and inconsistent clothing in AI comics. Many fans of Song history and military martial arts spontaneously produced secondary content, drawing character relationship diagrams and battle maps, and comparing the historical facts and fictional deductions in the plot.

The audience does not need to give up logic for the sake of pleasure. This is the core difference between it and the massive assembly line short dramas.

A creative team of six people

There are only six core creatives supporting this tens of billions of plays, and they belong to a loose creative cluster collectively known as “Old Eleven Theater”. The entire cluster has a total of more than 30 people, scattered in Guangdong, Anhui, Hubei, Jiangxi, Fujian, Hebei and other places. So far, most of the members have never met offline. The team does not have the fixed office space, attendance system, or fixed base salary of traditional film and television companies. It is a typical online distributed collaboration organization.

The origin of the team is said to come from an online AI course costing 1,000 yuan. At the beginning of 2025, a group of amateur students signed up for the same AI film and television training course, practiced together, polished each other’s works, and relied on external AI short drama contractors to work together and practice, gradually forming a stable collaboration circle. The members originally had a variety of occupations, including mothers, freight wholesalers, online ride-hailing drivers, and ordinary office workers. None of them had a professional film and television director or production background.

54-year-old Zhang Hansheng is the founder of Old Eleven Theater and the chief planner and IP person in charge of “Armed into the Song Dynasty”. His online name is “Old Eleven”. This name comes from the fact that he is the 11th student of the online course “Xuebao Studio”. He is from Jingxing Mining District, Shijiazhuang, Hebei Province. He used to be a freelancer. He experienced many entrepreneurial failures and was once burdened with debt. The turning point came when a change occurred at home and his father became ill. He chose to return to his hometown to take care of him. During his spare time, he read a lot of online novels and gradually came up with a simple idea to adapt high-quality online articles into visual short plays.

In March 2025, Zhang Hansheng officially began to learn AI video generation technology by himself, investing more than ten hours every day to study prompt words, models, and film production logic. In the beginning, he relied on accepting external short drama production orders to obtain income. When the market conditions were good in the early days, the production quotation could reach up to 3,000 yuan/minute. With the rapid iteration of AI-generated video models and the surge in supply, the unit price of the same type of production dropped to 200-300 yuan per minute in 2026. The profits of the OEM model were quickly compressed, which also forced him to turn to independent IP creation.

The first two short plays he completed independently were released one after another. The first part had over 10 million views, and the second part had over 100 million views and more than 300,000 collections on the platform. However, it was removed from the shelves 6 days after it went online because it hit the review red line. This experience made him more careful in selecting subjects and controlling the scale. The size of the team also expanded with the project, from just one person at the beginning, to three people on the third project, to five people on the fourth project, until “Armed into the Song Dynasty” was finalized as a core creative team of six people.

Zhang Hansheng positions himself as the “topic selection gatekeeper” of the entire team. His core job is to screen Internet IPs, determine the general direction of creation, and control subject matter risks. There were rumors circulating on the Internet that the team of more than 30 people, mostly unemployment fund analysts, were concentrated in a community in Xinxiang, Henan Province. Zhang Hansheng specifically came out to refute the rumors, saying that there were no fund analysts in the background of the team members, and everyone was scattered across the country to collaborate online. Now, members of the team have settled in the Short Drama OPC Creation Center in Shijiazhuang Jingxing Mining Area, and the local government has provided support policies of rent-free office space for the first two years.

Partnering with Zhang Hansheng and serving as the core of the team’s quality and process is chief producer and chief screenwriter Li Chaohua. Public channels have not yet disclosed his place of origin, education, and pre-entry occupation. It can only be judged from team interviews that his age is roughly in his thirties. He is a deep user of online articles and has been reading various works on Tomato Novels for a long time. His initial creative intention was to transform written stories that he liked and moved him into video content.

In the six-person creative team, Li Chaohua is the overall scheduler and final reviewer of the project. The division of labor among the six people is not as clear as a traditional crew. Each person is responsible for the corresponding paragraphs independently, dismantling the original work, rewriting the script, writing prompts, generating scenes, editing into the film, and finally submitting it to the group for cross-examination. The final decision whether it can pass the review and whether it can be included in the feature film lies in Li Chaohua’s hands. He set a strict internal standard. Even if the shot is smooth and has no obvious distortion, and only reaches 60 points, it will be directly judged as a waste card within the team and needs to be regenerated and polished.

“Polishing the script and lines is the most efficient way to save computing power and draw cards.” This is a point that Li Chaohua repeatedly emphasized. In his view, AI generation is just a tool, and the logic of the early text, character lines, and plot rhythm are the foundation that determines the minimum limit for the finished film. After “Armed into the Song Dynasty” became a hit, he pushed the team to make a counter-intuitive decision to reduce production. Prior to this, the team had up to five groups operating in parallel, with a monthly production capacity of up to 20 assembly line short plays. After the popularity, the team took the initiative to reduce production capacity to single digits. The computing power token consumption invested in a single drama increased, but the overall computing power consumption decreased, saving a lot of time to return to the original study, plot deduction, and polishing details.

He once talked about the team’s early living conditions, “Our family conditions are not very good. At the beginning, we could only survive by taking on popular dramas and make money first to stabilize the team. However, it is difficult to take into account the content quality of fast-moving production lines.” There are also loose teams that collaborate online. It is a warm bond that belongs to ordinary people. Members will send each other specialties from their hometowns, order cakes collectively for members’ birthdays, and send red envelopes to each other on the monthly settlement day. This bond has supported this group of amateurs through a long period of trial and error with almost no rewards.

Among the six main creators, He Xiushan serves as the chief director and is responsible for model debugging, picture style control, and prompt word system optimization. He is the most technically capable member of the team. There is no fixed salary for the entire team, and income is distributed according to work. Remuneration is settled based on effective working hours and labor division weights. Wasted films are not included in the share. The more work you do, the more you get. No online meetings unless necessary. Remote collaboration relies on WeChat groups, network disks, and online conferencing tools. The team originally planned to organize an offline party, and the location had already been decided. However, due to the sudden popularity of “Armed into the Song Dynasty” and everyone being caught up in the rush, the party plan has been put on hold.

The related company of the production-related entity “Jinye Theater” is Xinxiang Jinye Media Co., Ltd., which was registered in February 2026 with a registered capital of 500,000 yuan. Zhang Hansheng clarified the rumors of tens of millions of profits circulating in the market. He said that the project income has not yet been fully calculated. The returns of tens of millions reported online are external estimates, and it is not convenient to disclose specific revenue figures. He only admitted that the economic situation has improved significantly compared with the trough stage of entrepreneurship, and the previous debts have been paid off.

無投流

Judging from the popularity curve, the popularity of this drama completely deviates from the standard script of short dramas that buy more. The team has repeatedly confirmed to the outside world that there is no paid streaming in the entire process. After the episodes are reviewed and approved, they are uploaded and distributed to various platforms simultaneously. No traffic advertising is purchased. The main creator admits that he is not familiar with the streaming method. When the first season was launched, the platform had only 6 to 7 million collections, which was a tepid level. It was not until the day after the launch of the fourth season that the series topped the Hongguo AI new drama list, with popularity exceeding 50 million. It was not until the ninth season was updated that the team truly realized that the series was out of the circle; the popularity of the tenth season reached a peak of 82.06 million.

Its traffic comes from the superposition of several layers of free flywheels. The first is the underlying logic of the platform. The Hongguo short drama adopts the IAA advertising free model. The core appeal of the platform is the length of user stay and long-term retention, rather than short-term monetization of a single episode. “Armed into the Song Dynasty” has a very high completion rate, tens of millions of collections, millions of reservations for the next season, and a large number of users continue to follow it. It is a rare customer retention content on the platform and continues to receive the natural traffic tilt given by the platform algorithm.

The second is the spontaneous spread of vertical circles. History and military enthusiasts are the earliest core audiences. This group of users is extremely sticky and actively produces textual research pictures and analytical short videos. It has formed a large number of free secondary creations in Xiaohongshu, Bilibili, and Douyin, and continues to divert traffic to the outside world. It will further expand out of the circle and receive spontaneous recommendations from self-media bloggers. Central media have successively published comments and paid attention to this differentiated short drama. The positive evaluation at the official level has in turn strengthened word-of-mouth, forming a positive cycle.

However, it must be objectively admitted that this path is not replicable. The total investment in computing power is about 1 million yuan, which is the cumulative investment of long-term iteration and trial and error of nearly a thousand episodes in ten seasons. It is not a one-time burn. Relying on the platform’s free web IP library, there is no high IP procurement cost, which is the basis for the grassroots team to get started.

AI時代

AI tools have leveled the technical threshold for short drama production. Ordinary people only need to subscribe to computing power and learn prompt words to generate comic drama videos. After getting this tool, the vast majority of players on the market chose to compete for speed, production capacity, investment and flow price, and fell into homogenization. The victory of the “Armed into the Song Dynasty” team is essentially the victory of five levels of cognitive judgment. This is an asset that cannot be bought with computing power. Before the original work was selected, it had less than 100,000 reads on the entire Internet, making it an unpopular Internet article.

Zhang Hansheng does not give priority to backend popularity data when selecting IPs. He focuses on two points. One is that the story frame rejects brainless system flow, and the other is that the author has solid research on the official system, social style, and character conversations of the Song Dynasty. Identifying this potential target among tens of thousands of homogeneous cool articles relies on the aesthetic judgment accumulated through long-term reading and has nothing to do with the AI ​​model.

Second, when his peers were pursuing “the faster, the better, the more, the better”, Li Chaohua took the lead in making the decision to reduce production, proactively gave up the production line bonus, refused to mass-copy similar historical time-travel themes while they were hot, and restrained the impulse to realize short-term profits. To do addition, you need tools, and to do subtraction, you need concentration.

Third standard cognition. AI generates a qualified picture and a rough picture, and the computing power cost is almost the same. The quality red line of “60 points is a waste card” was drawn by the team itself. AI will not automatically bring about aesthetics, but will only infinitely magnify the gap between the creator’s own aesthetics and standards.

Fourth, the long-term serialization rhythm of one season per week and hundreds of episodes per season is matched with the repeated polishing of the script and picture of each episode. The upper limit of production capacity is given by technology, but when to update, to what extent, and when to slow down are the choices made by people.

The last thing is understanding the audience. During review, He Xiushan mentioned that before the series became popular, the team underestimated the market’s acceptance of heavy, restrained, and logical historical themes. The industry has long had a stereotyped perception that users of short dramas only need strong reversals and low-threshold instant pleasure. This drama proves that there are audiences in the vertical circle who are willing to accept complex narratives and are willing to be patient and pursue long-term stories.

At the same time, the team also remains clear-headed and refuses to create the myth that “AI can make ordinary people get rich easily”. Zhang Hansheng has reminded the outside world many times not to mistakenly think that learning AI to generate video will make a hit short play. He has personally seen a large number of companies entering the AI ​​short play track lay off employees and go bankrupt. He Xiushan is also worried that the cases will mislead the public and repeatedly emphasizes that there is a huge element of luck behind the hit. Zhang Hansheng broke down success into three elements: luck, choice, and hard work, with luck ranking first.

Innovative “Amateur Time”

By now, everyone is very clear about the ledger of the AI ​​short drama industry. AI has greatly reduced the costs of shooting and production, but the cost of traffic bidding continues to rise. Traffic often accounts for more than 70% of the total project expenditure. The settlement unit price of 10,000 effective plays continues to decline. Many projects are hard to produce, and the profits eventually flow to the traffic platform. The track is clearly differentiated. One type of players continues to follow the route of buying and selling, pursuing short, flat and fast ROI, focusing on standardized dramas with strong conflicts and low thresholds. The other type of players turns to vertical high-quality, long-term IP, hoping to rely on word-of-mouth and natural traffic to accumulate long-term IP assets. The latter has a higher ceiling, but is also more uncertain.

In this context, the greatest value of “Armed into the Song Dynasty” is not to prove that it can become a hit with zero investment, but to provide a precious control sample. What AI lowers is the threshold for getting started, not the threshold for success. Computing power, models, and tools can all be purchased with money, but the vision for topic selection, quality standards, determination of choices, and understanding of the story and audience cannot be purchased through human cognition.

Of course, the risks of this sample cannot be ignored. “Armed” now seems to be an isolated case, which is difficult to refine into a standardized methodology; and long-term serialization naturally faces pressure from weak plots and declining quality. At the same time, the copyright boundaries of AI content, review standards for historical themes, and the temptation of commercialization under the influence of popularity are all problems hanging over the team’s head. After the explosion, the team did not rush to copy the same type of historical themes, but turned to preparing new IPs with modern themes. When conditions are ripe, it will even try AI mid-length dramas and movies. Whether this road can be successful will take time to verify.

Li Chaohua added a very crucial sentence in the interview: If it had not been for the past year to undertake commercial orders and practice on dozens of short plays that no one cares about, there would not be a six-person creative team with a tacit understanding of each other, let alone the birth of “Armed into the Song Dynasty”.