In the autumn of 2026, a short drama called "Armed Into Song" ran through ten seasons on the Hongguo platform, carried not by a studio but by six people who had never worked in film, scattered across provinces and bound together through chat groups and cloud drives. Their story belongs to a familiar human pattern: ordinary lives such as driving, delivering, selling and raising children meeting a new tool and discovering that access is not the same as craft. Their leader insists that the tools have made the making cheaper but not the judgment easier, and that the audience, in the end, still ask
Six Remote Amateurs, One AI Short Drama: How 'Armed Into Song' Rewrote Expectations for Machine-Made Content
AI can lower costs, but lower costs do not mean lower standards.
Mark
Twelve billion views from six people working from home. Why does this one matter?
Mimi
Because the team itself says it wasn't the tools alone. They tested prompts repeatedly, and they judge a clean, error-free clip as only 60 out of 100.
Luke
Careful with the 12 billion, though. That comes via 36Kr, and it's cumulative across ten seasons, not a single audience.
Mark
And the 10 million yuan return?
Mimi
Zhang calls it a media estimate and says he didn't receive that. The cost figure, about 1 million yuan, is his own account.
Luke
Right, and that's self-reported. We don't have independent accounts.
Mark
What changed after the success?
Mimi
They went from about 20 dramas a month across five crews to single digits, and dropped assembly-line work.
Mark
Why does Zhang warn against the grassroots narrative?
Mimi
He says access to models is equal, but story judgment and editing skill are not.
Luke
Worth noting critics question the AI faces and the length padding. Whether the Northern Song setting holds up over time is unknown.
Il Polso
- Twelve billion views accumulated across ten seasons, yet the team's own leader publicly corrects the widely repeated profit figure, creating tension between myth and account.
- A six-person remote crew of former drivers, delivery workers and homemakers disrupts assumptions about who can make high-density historical drama.
- Viewers dissect official robes, travel permits and seating order in real time, pressuring the creators to match their pace with scrutiny.
- The team has cut back from five crews and about twenty dramas a month to single-digit output, choosing quality over volume.
- Its leader warns that AI democratizes access to models, not story sense, and the team is now preparing new intellectual property to test whether the method holds.
In the autumn of 2026, a short drama called "Armed Into Song" ran through ten seasons on the Hongguo platform, carried not by a studio but by six people who had never worked in film, scattered across provinces and bound together through chat groups and cloud drives. Their story belongs to a familiar human pattern: ordinary lives such as driving, delivering, selling and raising children meeting a new tool and discovering that access is not the same as craft. Their leader insists that the tools have made the making cheaper but not the judgment easier, and that the audience, in the end, still asks whether anyone truly felt what they put on the screen.
Across ten seasons from late October into early November 2026, an AI-made short drama called "Armed Into Song" grew from a modest debut into a phenomenon on the Hongguo platform, with 12 billion cumulative views and a trending topic on Weibo. Its makers are six remote creators with no film background, who once drove cars, delivered goods, sold wholesale or stayed home with children. Their head planner, Zhang Hansheng, began learning AI filmmaking only in March 2025, after a 1,000-yuan online course, and the team took its name from the studio he joined as its eleventh student.
Zhang is blunt about money. The widely repeated figure of a 10 million yuan return, he says, was a media estimate; the real cost was about 1 million yuan, mostly for computing power, and the team never buys traffic. Each member handles the whole process alone, from reading the novel to editing, and they rarely meet, relying instead on a year of rapport. Their internal standard is demanding: a clip without errors can still be "just very flat," and the funeral scene after a beloved character's death only worked, Zhang insists, because the person making it cried.
Viewers noticed the care. Robe colors, footwear and ceremonial seating were checked, and an online community built a research space to sort fact from fiction. The team has since scaled back, moving away from the "assembly-line dramas" they once made to survive, and is preparing new work. Zhang's warning remains: equal access to AI tools does not equal equal storytelling skill.
Ten consecutive seasons of an AI-made short drama, "Armed Into Song," ran on the Hongguo platform from late October to early November 2026. According to 36Kr, they drew 12 billion cumulative views. A single season peaked at a heat score of 82.06 million, and the show collected 11.41 million cumulative saves across seasons. The first season, launched in August, did only modestly. Popularity climbed with each later season, the fourth topped Hongguo's chart of new AI dramas, and on November 4 the Weibo topic "Hongguo's historical short dramas are a bit unexpected" entered the trending list.
The people behind it are six creators scattered across different provinces, working remotely. Their previous jobs included ride-hailing driver, small-truck delivery worker, wholesaler and a stay-at-home mother. None has a film or television background. Head planner Zhang Hansheng, 54, only started working with AI film production in March 2025, after buying a 1,000-yuan (about $150) online course. He became the 11th student of a certain studio, which gave the team its name, Laoshiyi Theater, or Old Eleven Theater.
Zhang is careful about the money. He says the widely repeated "10 million yuan" return (about $1.5 million) was a media estimate, is inaccurate, and that he did not receive anything like that. Costs were roughly 1 million yuan (about $150,000), mostly computing power. The team runs as a one-person-company model with no personnel costs, and the copyright was free. He also says the team never buys traffic: once a drama is finished and passes review, it goes out on several platforms at once.
Each member handles the whole chain alone: reading the novel, breaking down the script, writing prompts, generating video and editing. Work is gathered through WeChat groups and cloud drives for review. Visual consistency is largely handled up front through an asset library of character designs, scene images and props. Consistency of personality and emotion, Zhang says, cannot be solved with one standard face. The team rarely holds meetings, since some members work nights and sleep days, and they lean on rapport built over more than a year together.
The main tools are Seedance 2.5 for video and Doubao for text. Chief director He Xiushan says version 2.5 raised multi-image reference from nine images to more than ten, which helps with ensemble and large-scale scenes. Still, the creators decline to credit the model alone for the better look of later seasons; character, scene and lighting prompts were tested repeatedly. Their internal bar is blunt: a clip with no continuity errors and competent acting and camerawork might score only 60 out of 100, because it is "just very flat." An ordinary shot takes about three generations, battle shots often more than ten. For the funeral after Qin Hongying's death in battle in season six, He wrote a song and told Zhang he cried while making it. Zhang's view: if you don't cry yourself, the audience won't.
The source novel came from the copyright library in Fanqie's short-drama creator center. Zhang says it had fewer than 100,000 clicks then. He liked its passionate, hardcore tone and sense of homeland, which suit the male-oriented genre the team does well. It is a time-travel story without a "system": guns jam, ammunition runs out, and characters face household registration, money, grain and the military system. The author, "Chagen'er," is an English teacher.
Viewers noticed the density. Official robe colors, accessories, footwear, travel permits and ceremonial seating order were checked, some viewers shared Song dynasty history in danmaku comments, and netizens built a research platform labeling official systems and battle maps as "historical fact, fiction and speculation." Ban Yue Tan's commentary "This Short Drama Is Different" was reposted by CCTV.com, and Economic Daily and Guangming.com ran their own pieces. The shared thrust: AI can lower costs, but not standards.
The business has changed. The team once ran five crews and made about 20 dramas a month, some of which Zhang calls "assembly-line dramas" made to survive. Output is now in single digits, with no more such dramas. Novel reading and script polishing happen before generation and use no quota, so total computing use may fall even if a single drama costs more. Members are paid by division of labor and revenue share, and those paid by volume are counted on effective duration used after review. Jingxing Mining District in Shijiazhuang offers office space rent-free for two years, but Zhang does not plan to move the online process offline.
Zhang worries that "ordinary people can do it too" stories mislead people into thinking a class and some tools reproduce this result. AI democratization, he says, means equal access to models, not to story judgment, character understanding or editing skill. Critics have questioned the "AI faces," padding across a thousand episodes, and whether the Northern Song setting can sustain the story. Zhang concedes the team underestimated audiences, who accept complex, logical content. New IP is in preparation, and the six return to their own computers to read, revise, generate and send clips to the group chat for critique.
Citazioni salienti
Quality first, speed second.— Zhang Hansheng, head planner
If you don't cry yourself, the audience certainly won't cry.— Zhang Hansheng, head planner
Vuoi la storia completa?
Leggi l'originale su BigGo Finance ↗
Technology