Inside the 1,000-Person “AI Comic-Drama” Factory: Wild Growth and Content Ideals | A Conversation with 酱油
Summary
- The AI comic-drama boom is not AI taking over all creative work; it is short dramas having trained audiences to expect fast-paced consumption, combined with generated visuals clearing the “obviously AI” threshold. 酱油 says viewers have gone from “one episode a week” to having dozens or even hundreds of series to choose from each day; topic selection and scripts still determine hits: “Even if you make good content a little worse, plenty of people will still pay for it.”
- This is a labor-intensive content-production business. 酱油’s team expanded to 1,000 people in under 3 months, producing more than 100 series and roughly 6,000–8,000 episodes a month; a generation operator can produce about 10 minutes of material a day, while an editor can complete only about 1–2 minutes, making post-production the largest workforce: “Post-production cannot be replaced by AI; it can only be done by hand.”
- By 酱油’s own figures, the company is on track for annual revenue of about RMB1B and profit of RMB200M–300M; it does not keep quality spending down through corner-cutting. Each series costs about RMB200K–300K, while its in-house cost is roughly RMB1,500 per minute versus a market average of RMB1,000; most companies generate 40 images per scene, but his team requires more than 160, believing investment in quality matters more.
- The core asset is the human judgment required to compress novels of millions of characters into visual stories, while AI currently focuses mainly on image and video generation and reading efficiency. A roughly 120-minute comic drama carries about 60,000 characters, while source novels often run 1M–5M characters; AI may let a screenwriter distill and nearly finish a 2M-character novel in about 30 minutes, but it struggles to distinguish the essential from the incidental and cannot independently solve the adaptation problem of turning “a thousand readers, a thousand images” into one image seen by everyone.
- 酱油 divides the business into three product lines—commercial content, technical spectacle and globally legible IP—and uses AI to lower the cost of film-scale experimentation. The first serves the “stress relief” needs of 18-year-olds, the second plans a visual feast built around a god-and-demon war like The Battle of the Gods, and the third aims for universal themes in the vein of Coco; his math is simple: “At RMB200K–300K a title, 100 titles cost only tens of millions,” and one breakout topic could recoup the full investment with a profit.
- The 1,000-person expansion follows a “factory management” model, with ordinary employees and managers taking on different levels of intensity according to role and pay. Ordinary employees work 8 hours a day with a 2-hour lunch break, and team leaders and below do not work overtime; managers above team-leader level do, and the management team also works weekends. 酱油’s view is that “content drives production, not production driving content”: as long as the content works, replacing production staff is not a major risk.
- The financing goal is not to plug a current cash shortfall but to go public, expand overseas and become an application-focused AI concept stock that has “actually made big money using AI”; by contrast, he gives third-party AI comic-drama tool companies zero investment value. His reasoning is that tools depend on upstream models, while a model or tool company could replicate their functionality within 5 days; he expects one disruptive AI product after another in 2026, and thinks that within 3 years cinemas may be showing films with more than 10 minutes of AI-generated footage that audiences could not identify in advance.
Deep dive
1. AI Comic-Drama Demand Starts with Short Dramas; Visual Quality Is the Inflection Point
酱油, 29 and a vocational-secondary-school graduate, describes his work as “disrupting an era of visual media.” The host’s scorecard: after going all in this September, his team reached 1,000 people in under 3 months and gross receipts exceeded RMB50M; of the 22 comic dramas on Douyin that surpassed 100M views, 5 came from his team. 酱油 expects annual company revenue of about RMB1B and profit of RMB200M–300M, and forecasts the team will reach 3,000–5,000 people next year.
酱油 believes the demand shock was made possible by short dramas training users to expect fast-paced content. AIGC then expanded animated supply from “one episode a week” to dozens or even hundreds of titles a day, giving viewers far more content to choose from.
He sees the most direct technological breakthrough in visual quality: “With every video generated by AI today, you can no longer tell that it is AI.” In his view, only after image quality clears that perceptual threshold will higher output capacity translate into real market attention.
2. A 1,000-Person Team Runs on a Layered Production Line
A comic drama starts with topic selection and then moves through screenwriting, storyboarding, AI artists setting characters and locations, supervisors and team leaders breaking down tasks, generation operators producing material, and team leaders handing the work to post-production for editing, dubbing and release.
AI artists are not generation operators: the former lock in the male lead, female lead, supporting characters and settings, creating the shared asset package for the entire series; the latter generate effects, scenes and dialogue from text and are the lowest-barrier role—“if you understand text and know how to operate AI tools, you can do it.”
The technology team is not trying to build a foundational model. It is building dedicated workflows and tools for this organization. 酱油 says the only metric that matters is whether the system improves each employee’s productivity.
The company delivers more than 100 premium comic dramas a month, each running about 60–80 episodes, or roughly 6,000–8,000 episodes in total. A generation operator can produce about 10 minutes of material a day, while an editor can complete only about 1–2 minutes, making post-production the largest department.
3. AI May Let Screenwriters Grasp a Novel’s Structure in 30 Minutes, but Humans Still Decide What to Cut
酱油 explicitly limits AI’s role at the front end of creation: “Text still has warmth. What AI produces is too formulaic.” AI is mainly involved in image and video generation; scripts, editing and dubbing have not yet benefited much from it.
The adaptation problem is compression. Novels range from 1M characters to 3M–5M characters, while a roughly 120-minute comic drama is about 60,000 characters. If AI mechanically preserves every event in the original, it drags the filler down alongside the main plot: “You can no longer tell what matters and what doesn’t.” Humans decide what to cut and what must remain.
The deeper difference lies in the medium. A novel may describe “a full moon,” leaving each reader with a different image; a comic drama puts the moon directly on screen, so everyone sees the same thing. “A novel gives a thousand readers a thousand images,” while visual media must turn that into a concrete expression broad audiences can accept.
AI still improves screenwriter productivity: it can first extract the outline and story arc of the entire book, allowing a screenwriter to nearly finish a 2M-character novel in about 30 minutes before concentrating on how to adapt it, rather than spending the time reading every word. 酱油 believes the script remains the role that truly determines content quality.
4. Content Talent Is Scarce, and Visual R&D Determines Whether a Topic Is Packaged Correctly
酱油 considers screenwriters the hardest people to recruit, followed by R&D staff; screenwriters, R&D and technology are also the 3 highest-paid categories outside the founder. Generation operators, AI artists and post-production staff are more common and relatively easy to hire.
“R&D” here does not mean foundational technology development. It means continuously testing visual systems including photorealism, 3D, 2D, Japanese animation, American comics and Western fantasy. 酱油 believes subject matter and visual style must match: a romance novel is not suited to direct 3D treatment and may need “a 新海诚-style.”
When recruiting, 酱油 talks about money first: “To move someone from a familiar environment into an unfamiliar one, you have to win them over with money, not sell them a dream.” Only afterward does he discuss a shared vision—whether they are willing to be “countercurrents of the era” and build a new visual age together.
5. RMB200K–300K per Series, with Quality Spending Concentrated on People and Shot Density
The market typically charges about RMB1,000 per minute; 酱油’s team spends about RMB1,500. Total investment per series is roughly RMB200K–300K. By his account, labor represents more than 60% of costs, compute about 20%, scripts about 15%, and dubbing under 10%, at roughly 5%.
The direct reason for the higher cost is shot density: most companies generate 40 images per scene, while his team requires more than 160. 酱油 compares himself only with peers in this segment, targeting “the best image quality and the best visual effects,” without claiming to have surpassed Chinese animation or the world’s top animation studios.
Asked whether he cares about the prices of different models, 酱油 says cost matters, but not at the expense of quality: “Making a little more money is not very meaningful.” His long-term path is from AI comic dramas into AI films and AI animation, not to harvest the boom and then switch sectors.
The team mainly uses 生数科技, 可灵, 纪梦, Banana and Sora: it prefers 生数科技 simply because it likes the visual style; 可灵 is better suited to detail in large-scale scenes, while 纪梦 performs better on some image-generation tasks. The strategy is not to bet on one provider but to combine the strengths of all of them.
6. “AI Flavor” Is a Transitional Bias, and 酱油 Compares the Evolution to the Shift from Hand-Drawing to Computers
Market criticism often treats “obviously AI” and “too much AI flavor” as synonyms for shoddy work. 酱油 emphasizes that AI improves every day, saying, “Things made by AI have already surpassed 80% of the elite.” He believes low-quality supply in the sector has also created the impression that “anything AI makes is garbage.”
The host challenged AI content’s cheapness with the analogy of pre-prepared food. 酱油 rejected it because pre-prepared food can affect health, while AI-made animation “does no harm to the human body.” He does not think AI will replace top-tier “S-series” animation, but believes it can fill the underserved A-, B- and C-tier markets left by traditional production—and may even do them better.
Citing 宫崎骏’s insistence on hand-drawing and view that computer lines lack warmth, he frames the dispute as a generational transition: hand-drawing once resisted computers, and now computer production is resisting AI. “The transition period is the hardest time for other people to understand,” but he believes AI will eventually become the mainstream of visual production.
酱油 forecasts that over the next 10 to 50 years, about 50% of all visual content people can see may be generated by AI. He says this will not ruin a generation; it may instead give young people opportunities for wealth, status and upward mobility. He also cites a partner whose annual income had previously been below RMB300K but who later drove away in a Rolls-Royce after one year.
7. A Three-Layer Content Mix Brings Commercial Demand, Technical Spectacle and IP Ambition Together
The first product layer directly serves the commercial market, using straightforward, high-satisfaction content to help “18-year-old boys” decompress under the pressure of daily life. 酱油 says today’s comic dramas can be understood as “stress-relief videos.”
The second layer prioritizes technology and spectacle, resembling the The Battle of the Gods or Avatar in his mind: the story can be relatively familiar, but the scale and effects must create visual impact. The team plans to produce The Battle of the Gods, using AI to depict a large-scale war between gods and demons and defining the product as a “visual feast.”
The third layer is IP-driven content, benchmarked against Coco, using themes of death, life and social problems to create resonance. He believes Chinese film and television, as well as short dramas, have difficulty going overseas because their local cultural barriers are too high; simple stories like Titanic and Catch Me If You Can are easier for different markets to understand.
AIGC changes the cost of experimentation. 酱油 says developing a traditional film may require tens or even hundreds of millions of RMB, while an AI production may cost only RMB200K–300K; 100 titles would therefore cost only tens of millions. Low-cost experimentation means the team can keep trying even if some projects fail, and one breakout topic could recoup the investment with a profit.
8. From Writing Overnight in Internet Cafés to Short-Drama Windfalls, 酱油 Has Always Been in the Content Business
Coming from an ordinary family and lacking a computer, 酱油 paid for overnight sessions at internet cafés to write novels; the fee at the time may have been only RMB8–10. He quit gaming, peaked at 50,000 characters a day and sometimes went 24 hours without sleep. “For ordinary people like us, there may be only one opportunity in a lifetime,” so when a novel failed, he immediately moved on to the next.
He says he earned more than RMB10K in his first month in the industry and reached annual income of RMB1M within 6 months. He then founded offline web-fiction incubation classes, recruiting aspiring writers, teaching them to write, and submitting their work to platforms in exchange for a share of revenue. That business lasted about 8 years in web fiction but still did not make him fully financially free.
Short dramas delivered the wealth jump. At first, the maximum cost of one series was about RMB150K; later, RMB200K–300K could produce one. He says he earned more than RMB7M in his best month and more than RMB50M in total over a little more than a year; pretax income was above RMB50M, while after tax he may have taken home only slightly more than RMB30M. He also cites a friend at 山海集团 who started with roughly RMB2M in capital and, by his incomplete estimate, earned RMB300M–500M in a year.
Moving into comic dramas was not a departure from content but a response to short dramas evolving from a “low-cost, high-return” business into an entertainment-industry business dependent on powerful promotion, star effects and fan culture. “This is not making money through content,” he says, so he went all in on AI comic dramas, which require no actors and do not require him to manage actors’ moods, believing that good content that satisfies viewers is enough.
9. 酱油’s Talent Theory Is Ruthless: Effort Can Make Money, but Cannot Guarantee Success
On the relative weight of effort and talent, 酱油 gives an uncompromising answer: “People without talent are useless. No matter how hard they work, it won’t help; they are only wasting time.” A first piece of writing usually reveals the signal. Only if a writer shows a clear strength in story, prose, imagination or trend-following is further development worthwhile.
His own evidence is uneven ability: aside from exceptionally strong Chinese-language scores, he failed mathematics, English, physics, chemistry and geography. His advantage is not polished professional speaking but explaining things clearly through down-to-earth analogies. “1% talent matters more than 99% effort.” Effort is more likely to earn a boss’s recognition and higher income than success itself.
He compares AI comic dramas to “turning a landline into a mobile phone.” When Nokia appeared, the first thing it unexpectedly displaced was not the landline but the flashlight; the phone later also carried WeChat, the internet and video. AI likewise will not stop at comic dramas but will enter games, code, PowerPoint and many other yet-to-be-defined formats.
10. Factory Management Supports Rapid Expansion, with Intensity Allocated by Pay and Rank
酱油 denies that expanding to 1,000 people requires complex large-enterprise management, describing the core as “factory management logic”: supervisors and team leaders distribute tasks layer by layer, and employees leave once their work is done. The organization mainly consists of administration, ordinary employees and middle and senior management.
Team leaders and below do not work overtime: they are on site for 8 hours a day, including a 2-hour lunch break, for about 6 actual working hours. 酱油 says ordinary employees receive an annual base salary of roughly RMB50K–60K, which may work out to about RMB5,000 a month including social insurance. Having worked as a server, mover, delivery rider and property agent, he wants employees to retain lives, friends and families after work, and at least feel that “this company is actually pretty good.”
The other side is that team leaders and above must work overtime, while the management team also works weekends. 酱油’s allocation principle is equally blunt: higher pay means taking on more for the company. He agrees with “treating every day as the last day,” because no one knows when the boom will end.
His reason for continuing to work at high intensity is not consumption—his wife controls the money, he does not buy watches, and even his drinks cost “RMB1.50”—but the desire to build a brand, go public and even create a new AI concept stock. “Having ideals” is his explanation for why he has not stopped after reaching financial freedom.
11. He Rejects the Idea of Industry Rivals, Yet Wants to Raise Capital and Become an AI Application Concept Stock
酱油 says the content industry “has never had peers”: Under One Person, Fox Spirit Matchmaker, Crayon Shin-chan and Detective Conan may have similar styles or distribution channels, but their stories are irreplaceable. “He can never take what is in my head, and I can never take what is in his.”
The host pressed him on whether talent and paid acquisition still create competition. 酱油’s response is that technology will ultimately become equalized; leaders have only reached “level two” first. Talent moving between companies still means doing similar work, while paid acquisition ultimately depends on content: bad content only loses more money when promoted.
Although the company is already profitable, he still wants to raise capital and go public because he believes the AI companies discussed by the market are mostly chip, technology or graphics-card sellers. He wants to become an application concept stock that has “actually made big money using AI.” Financing and an IPO would help the company shake up the Chinese animation market, expand overseas, make games, and develop guzi culture, figures and designer toys around its own IP.
For third-party AI comic-drama tools, he gives an investment verdict of “none at all—zero value.” If upstream foundational models stop allowing access, tool companies could lose their foundation; even if that probability is low, it cannot be ignored. More importantly, he believes a relevant model or tool company could see that a tool works and spend 5 days upgrading its own product to match it exactly. In his analogy, tool vendors are merely intermediaries that write instructions for a shovel, package it and mark it up by RMB2.
He predicts that 2026 will bring one AI product after another that “upends everyone’s imagination,” and says the team already has a successful “top-secret” project that will be announced after the Lunar New Year. Over a 3-year horizon, he believes most visual content will be replaced by AI; AI will move beyond Douyin and Kuaishou onto a larger stage, and a theatrical film may already contain more than 10 minutes of AI-generated footage that audiences would find difficult to identify in advance.