“制作已经没有任何意义了。”
在近期举办的2026阅文创作大会结束后的交流环节，黄浩南发出如此感叹。
而最近让他这种感受更强烈的一件事，是他发现一个初三女生在校用豆包做了一个AI短视频，瞬间收获300万点赞。这在很多短剧公司看来，是烧光几百万甚至上千万元，都可能难以实现的成绩。
黄浩南是酱油文化创始人，网文作家出身，如今是国内体量最大的AI漫剧出品方之一。酱油文化可以做到月均上线两三百部剧。他还开启了海外布局，海外收入占公司大盘约三成，每月稳定一百多万美金。
猝不及防的是，今年AI正在侵蚀他的护城河。
DataEye研究院数据显示，2026年 1—5月国内 AI剧漫剧市场规模已达220亿元，全年有望突破400亿元，同比增长约138%。供给端的数字更惊人：上半年，抖音新增 AI剧、漫剧22.19万部，日均上新超过1200部。广电总局的数据披露，今年前8个月全国上线微短剧43万部，是去年全年的13倍，AI剧占比超九成。
供给不再稀缺。高峰期，大约每20秒就有一部新剧诞生。然而，上半年抖音新增的22.19万部AI漫剧中，爆款率仅0.47%，超九成内容流量表现低迷。
技术平权，并没有给创作者带来“黄金时代”，反而当供给无限膨胀，注意力成为终极稀缺品，整个行业的商业模式、竞争壁垒与价值分配，正在经历一场静默的底层重构。
从劳动密集到算力密集
黄浩南给出了一组关于AI漫剧成本重构的数据。
去年，行业做一部AI漫剧的标准价是10万元，每分钟1000元；到今年10月降到每分钟400元，一部80分钟的作品约4万元，下降约60%。
成本结构也发生了变化，主要体现在token占比上。
按黄浩南的估算，去年token在成本中的占比为20%至30%，大头在人工、剪辑与配音。而今年token占比达到了50%，人工占比为20%，剧本创作占10%至20%，剩下的是水电房租等开支。
阅文集团AI平台部负责人陈炜于侧面印证了黄浩南的说法。
陈炜于表示，年初人、算力、其他的比例大约是6:3:1；现在批量剧已降到3成以下，算力占6到7成。
也就是说，AI漫剧在一年内从劳动密集型生意变成了算力密集型生意。
随之而来的是，组织形态的剧烈收缩与效率的极限提升。而黄浩南公司的员工数，也从巅峰 1500人缩减到现在几百人。他强调，今年缩减公司规模的逻辑，不是因为公司的经营状况的问题，而是“因为不需要这么多人”。
AI已经使得酱油文化的产能，从去年10个人1个月可以做1部剧，变成今年1个人1个月可以做3部剧。而陈炜于观察到，行业内“批量剧”，甚至已经达到了1人1天产出1部的量。
然而，技术带来的效率红利并没有沉淀为生产者的利润。漫剧公司的利润并没有增长，一个重要原因是算力变贵了。而随着AI生产质感的提升，现在如果要做一部相对精良的短剧，还需要更贵的算力。
也就是说，当核心投入从人力变成token，看似在用AI替代人，但其实只是把一笔可谈判的人力成本（工资），换成了一笔难以谈判的成本（算力）。Token的价格由模型厂商决定，漫剧公司，尤其是中小型公司并没有议价权。
AI的进步，使得入局和退出变得更加容易了。而更低的入局门槛和退出成本，也意味着更多人可以无序地涌入这个赛道。
单价崩塌与概率游戏
供给无限的第一个后果，是单品价值崩塌。
黄浩南给了一个对照：“去年赚了一千万叫爆款，现在赚十万、二十万就叫爆款了。”
两个数量级的落差，发生在一年之内。这在今年下半年短剧出海市场上表现得更为明显。
黄浩南说，一开始，在TikTok上，他1亿播放量的剧，可以分账14万美元，5亿播放量分账12万美元，3亿播放量分账4万美元。
分账大幅下降的一个核心原因是，出海短剧供给端极速膨胀。他从TikTok那里获得的消息是，平台上新量， 6月份还不到1000部，7月不到 5000部，8月10万部，9月估计20万部。短短三个月，从蓝海到红海，“原因就是太卷了”。
播放量疯狂攀升，收益持续下降，单位播放收益跌去八成以上。
行业数据也支持黄浩南的判断。DataEye研究院称，今年10月，AI漫剧万播分账已从早期的30元至100元回落至5至10元。
上半年抖音新上线的22.19万部作品中，播放破亿的仅1055部，破亿率0.47%；以 5000万播放为盈亏平衡线，达标率不足1.3%。证券时报测算，约98.7%的项目上线半年内无法回本。
这也解释了黄浩南为什么坚持走量。他的公司，原来月产四五十部剧收入在两三千万，现在月产两三百部剧收入仍是两三千万。
同样的收入，需要五倍产能去换。也正因如此，黄浩南对自身公司的多元化发展不抱希望，这与大多数头部公司分散单一收入来源风险的做法背道而驰。
他坚持表示，不做APP，不做平台，因为他看到了他朋友因此亏了两个多亿；他也不考虑做AI电影等周期跨度长的高风险商业项目，认为那是属于传统影视巨头的游戏。
供给无限，经典归零
如果只是钱难赚，这可能是一个产业初期泡沫的现象，或者，一个普通的产能过剩故事。黄浩南真正看到的是，AI摧毁的不是利润，而是记忆点。
“AI从出来的第一天开始，就把成名之路给封了。”他说。
在传统影视时代，一部《铁齿铜牙纪晓岚》可以流传数十年，成为几代人的集体记忆。但在AI时代，哪怕做出一个同样级别的东西，也流传不了多久。
AI不是不可以生产高质量内容，但是，内容供给的极度膨胀，摧毁了“经典”赖以生存的稀缺性土壤。
他举了自己儿子的例子，孩子只对“超级飞侠”和“汪汪队”有感情，因为他只能看这两部。如果今天让他看几百部，他也记不住。
当一个月有30万部作品产出，其中爆款300部，一天爆10部，观众的注意力被彻底稀释，任何单一作品都难以形成持久的记忆锚点。
就如文章开头他提及的那个收获300万点赞的中学生，在AI短视频算得上是一个不错的成绩，但是很显然，明天她未必做得出第二个，也没有人为她的下一部买单。而因哪吒系列动画电影出名的导演饺子，下一部电影五到十个亿的票房成绩大概率是有保障的。
AI让“超级个体”、“天才”越来越多，甚至每个人都可以成名一天，就像抽奖一样，今天你爆，明天他爆，但这个人是谁，我们记不住，也不重要。
这样一来，IP很难再从AI内容里长出来，且“爆款”的窗口期很短，“经典”这条路不可逆地关闭了。而内容产业的一部分商业价值，很长一段时间，建立在一个好IP可以被反复变现的路径上，而在供给无限的环境里，这一模式就要失效了。
作为坐拥大量IP的平台方代表，陈炜于认为，AI时代IP的多样性变得重要，“内容生态是一个‘面’的生意，不是一个点或者一条线的生意。”
“未来的优势都是阶段性的优势，不能说所有的优势。IP储备、制作能力、平台渠道、算力成本，这些要素的重要性会随着技术演进和市场变化而动态调整。唯一不变的，是能否讲好这个故事，能否研究好这个生意本身。”陈炜于说。
机构的“黄昏”：权力走向两端
行业在规模上空前繁荣，在价值沉淀上，却面临前所未有的困境。AI剧这门生意的形态被迫变异，短漫剧公司始终在跟着市场跑。
今年1月，黄浩南花了很大精力在做AI仿真人剧，理由很实在：红果给了政策，“把最近爆过的剧随便改一点，甚至不改直接放上去，都是拿麻袋在装钱”。
几个月后，他转向了3D剧，因为“所有人都做仿真人了，这个生意不好玩了”，而做3D剧还有个逻辑是，孩子们放暑假了，暑期动漫是旺季。
“没办法，前几年是老天爷‘喂饭吃’，现在是求着老天爷、抢饭吃。”黄浩南说。
当作品的生命周期，以周甚至是天来计算，在这个模型里，创作者卖的不是作品，是用户时长。
黄浩南认为，短漫剧就是打发时间，“让你三个小时坐在这里看手机，成了，平台成功，我成功，你成功。”
今年7月，第三方平台统计，红果日活达到1.68亿，超过腾讯视频、爱奇艺、优酷、芒果TV四家之和；人均单日停留125分钟，超过抖音的112分钟和微信的82分钟，引发全行业热议。
不过，平台消费的，很大程度上，不是“作品”，而是广告库存。在那里，作品只是用户停留时长的载体。
强大的虹吸效应，使得90%以上的短漫剧机构需要依靠红果生存，10%的人靠其他渠道赚钱。“没办法，红果日活太强了”——这是他们普遍的感叹。
在淘金热里卖水始终是好生意，但最好的生意是拥有金矿，或者拥有那条河。
渠道的议价权正在上升。8月27日，红果调整AI剧的剧本合作政策，S+和S级保底直降 75%，A+和A级取消保底；12月底前全面下线“看剧赚金币”模块。
补贴退潮，风险下移。保底大幅缩水或取消后，原本由平台承担的剧本收益不确定性部分转移给了创作者和承制方，创作者需更多依赖分账表现来获取回报。金币功能下线，也被视为平台减少现金补贴拉新、转向留存核心用户和广告变现效率的信号。
一个风口，因为AI的加持，门槛低到人人可入，它就不再是一个风口，而是一次权力的转移分配。黄浩南预言，在2026年结束之后，中国将没有AI漫剧机构，不再有“机构”这个逻辑了，未来全是OPC。
当OPC成为越来越普遍的组织形态，内容制作的产业结构，也会从“机构中心”向“个体中心”迁移。收益属于那些跑得快的人、运气好的人，以及，卖铲子和收过路费的人。
当“天才”不再值钱，“经典”难以诞生，“爆款”越来越像一场概率游戏，内容产业的参与者们或许需要重新思考的是：究竟在为什么样的价值而生产？（本文首发于钛媒体APP，文｜LccBeta）
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
"Producing is no longer meaningful."
Ho Nam Hwang lamented the exchange that followed the recent conference on reading and writing.
One of the things that made him feel more strongly recently was that he discovered that a schoolgirl had made an AZshort video out of Doubao and had a quick gain of 3 million. This, in the view of many short play companies, could be difficult to achieve by burning millions or even millions.
Huang Ho Nam, the founder of the soy sauce culture, is an Internet author and is now one of the country ' s largest Ai comics. The culture of soy sauce makes it possible to go online 200 or 300 plays per month. He also opened an overseas layout, which accounts for about 30 per cent of the company's revenue, stabilizing over $1 million per month.
This year, AAI is eroding his moat.
Data from the DataEye Institute show that the market for the AAI drama in the country reached 22 billion yuan in January-May 2026 and is expected to reach 40 billion yuan throughout the year, an increase of about 138 per cent over the same period. The supply-side figures are even more striking: in the first half of the year, Douyin added 229,000 new AI plays, comics and more than 1,200 new daily. According to data from the General Directorate of Radio and Television, there were 430,000 online micro-shortcasts in the first eight months of the year, 13 times the number for the entire year, and the number of AI plays was over 90 per cent.
Supply is no longer scarce. At the peak, about every 20 seconds a new play is born. However, in the first half of Douyin ' s 22.19 million new AI comics, the rate of bursting was only 0.47 per cent, and over 90 per cent of content flows were underperforming.
Technological parity, instead of bringing about a “gold age” for creators, is going through a silent, bottom-up re-engineering of the business model, competitive barriers and value distribution of the entire industry, when the supply expands indefinitely and attention becomes the ultimate scarcity.
From labour to energy.
Huang Ho Nam gave a set of data on the cost of the Ai Comic re-engineering.
Last year, the industry standard price for an AI opera was $100,000 per minute; by October this year it had fallen to $400 per minute, or about $40,000 per 80 minutes, a decline of about 60 per cent.
The cost structure has also changed, mainly in the token ratio.
According to Huang Ho Nam ' s estimates, last year Token ' s share of costs ranged from 20 to 30 per cent, with large numbers of manual, clipped and voiceover. This year, token ' s share was 50 per cent, manual 20 per cent, script creation 10 to 20 per cent, and other expenses such as electricity and water rental.
The statement of Hoang Ho Nam was corroborated by Chen Zheng, head of the Ai Platform Department of the Reading Group.
Chen Zheng indicated that the population, calculus and other ratio at the beginning of the year was about 6:3:1; now the mass drama has dropped to less than 30, or 60 to 70.
In other words, in one year, the AI Comics changed from a labour-intensive to a math-intensive business.
This was followed by a sharp contraction in organizational patterns and an increase in efficiency. And Huang Ho Nam's number of employees has shrunk from top 1,500 to hundreds today. He stressed that the logic of reducing the size of the company this year was not a matter of the company's operating conditions, but “because so many people were not needed”.
AI has enabled the production of soy sauce culture to move from one in 10 people last month to three in one person this year. Chen Zheng observed that “volume dramas” within the industry had even reached one person per day.
However, the efficiency dividend generated by technology does not sink into the profits of producers. The profits of Comics have not increased, and one important reason is that they have become more expensive. And as the quality of AI production rises, it will now be more expensive to do a relatively good short play.
In other words, when core inputs move from manpower to to token, it seems to be replacing them with AI, but it is simply a negotiable human cost (wage) in exchange for a difficult cost (calculation). Token ' s prices are determined by model manufacturers, and comics, especially small and medium-sized companies, have no bargaining power.
Progress in AI has made entry and exit easier. Lower entry thresholds and exit costs also mean that more people can enter the track in an irregular manner.
Unit price collapse and probability game
The first consequence of unlimited supply is the collapse of individual value.
Hwang Ho Nam contrasted: "The $10 million he made last year was called "black money," and now 100,000, 200,000 is "blow money".
The difference between the two orders of magnitude occurred within one year. This was even more evident in the second half of this year in the short-time out-of-sea market.
Huang Ho Nam said that, at first, on TikTok, his 100 million-dollar play could be split between $140,000, $500 million by $120,000 and $0.040 million by $300 million.
One of the central reasons for the sharp decline in the distribution was the extremely rapid expansion of the supply end of the short-run drama. He got news from TikTok that there were less than 1,000 new ones on the platform in June, less than 5,000 in July, 100,000 in August and an estimated 200,000 in September. In just three months, from the Blue Sea to the Red Sea, “the reason is too thick”.
The number of airtime shows is rising wildly, the returns are falling, and the unit's airtime is falling by more than 80 per cent.
Industry data also support Huang Ho Nam's judgment. According to the DataEye Institute, in October of this year, the Ai ' s Mandarin sub-counting dropped from an early $30 to $100 to $5 to $10.
In the first half of the 22.19 million new Douyin entries on the line, only 1,055 of the 10 billion pieces were played, with a rate of 0.47 per cent, and the 50 million was shown as a balance line, with a rate of less than 1.3 per cent. The securities are measured at the time that about 98.7 per cent of the projects cannot be returned within six months.
This explains why Huang Ho Nam insisted on going. His company, which originally produced 40 and 50 plays per month, had a revenue of 20 and 30 million, and now has a revenue of 20 and 300 plays per month.
The same income requires five times the capacity to replace it. This is also why Ho Nam Huang is not hopeful about the diversification of his company, which runs counter to the risk of most head companies spreading a single source of income.
He insisted that he did not do the APP, did not do the platform, because he saw his friend loses more than 200 million; he did not consider doing a high-risk business with a long cycle, such as an AI film, which he considered to be a game of traditional visual giants.
The supply is infinite. The classic is zero.
If money is hard to earn, it could be an early bubble in an industry or a common overcapacity story. What Huang Ho Nam really sees is that AI destroyed not profit, but memory points.
"Ai closed the path of fame from the first day out." He said.
In the age of traditional video, an " iron and copper tooth " can be passed on for decades as a collective memory for generations. But in the AI era, even one thing at the same level will not pass for long.
AI is not incapable of producing high-quality content, but the extreme expansion of content supply destroys the scarce soil on which the “classic” depends for survival.
He gave the example of his son, who only had feelings for the “Super Flying Man” and the “Winds” because he could only watch them. If he'd seen hundreds today, he wouldn't remember.
When 300,000 works are produced a month, 300 of them explode, 10 explode a day, the audience ' s attention is completely diluted, and no single work can create lasting memory anchors.
As he mentioned at the beginning of the article, the high school student who received 3 million points of appreciation was a good achievement in AZshort video, but it is clear that tomorrow she will not have to do a second and no one will pay for her next. And the next movie's five to a billion-dollar box office is probably guaranteed because of what's famous for animated movies.
AI makes "super individuals" and "genius" more and more and more, and even everyone can be famous for a day, like a lottery, today, and tomorrow, but who is this guy, and we can't remember and it doesn't matter.
As a result, it is difficult for IP to grow out of AI content and the window for “explosion” is short, and the “classic” road is irreversibly closed. A part of the commercial value of the content industry has been built for a long time on a path where good IPs can be converted over and over again, and in an environment where there is no limit to supply, this model is going to fail.
As the representative of a platform with a large number of IPs, Chen Zheng believes that the diversity of IPs of the AI era has become important, “the content ecology is a `face' business, not a dot or line business”.
“The advantages of the future are of a phased nature and not all. The importance of IP reserves, production capacity, platform channels, and power costs will evolve as technology evolves and markets change. The only thing that remains is whether the story can be told and whether the business itself can be studied.
Agency “Dizzy”: power to both ends
The industry is experiencing unprecedented prosperity on a scale and values, while facing unprecedented difficulties. The shape of the Ai drama business had to be changed, and the short opera company was always following the market.
In January of this year, Hoang Ho Nam spent a great deal of effort on the Ai simulation of the human play, for a very good reason: the red fruit gave the policy that “replaces, or even puts directly, the latest blasts with sacks of money”.
A few months later, he turned to the 3D show, because “Everyone's doing it, and the business is not funny”, and the logic of doing 3D is that the children are on summer vacation and the summer season is a good season.
“There is no other way to do it, but in the last few years it was God who `feed' and now begs God to eat.” Huang Ho Nam says.
When the life cycle of a work is calculated by week or even day, in this model the creator does not sell the work, but the user's time.
Hwang Ho Nam believes that short opera is about time, "You sit here for three hours and watch your cell phone."
In July this year, the third-party platform counted 168 million red fruit per day, more than the sum of Tencent videos, Achilles, Yocheng, Mango TVs; 125 minutes per person per day, more than 112 minutes per person per day and 82 minutes per person per day of Douyin, and 82 minutes per person per day of WeChat, which provoked industry-wide enthusiasm.
However, the platform consumes, to a large extent, not “works”, but advertising stocks. There, the work is just a long-stayer carrier for the user.
The strong rainbow insinuation has made over 90 per cent of the short opera institutions dependent on mangroves for their survival, and 10 per cent of people rely on other sources for their income. "No way, red fruit lives too hard" — that's their general lamentation.
It is always a good business to sell water in gold-mining heat, but the best business is to own a gold mine or the river.
Channel bargaining power is rising. On 27 August, the S+ and S levels were lowered by 75 per cent and the A+ and A levels were eliminated; the full-scale drop-off of the “see the play for gold” module by the end of December.
Subsidies retreat, risk shifts downward. The uncertainty of the benefits of the script, which would otherwise have been borne by the platform, has been shifted in part to the creators and the sponsors, who need to rely more on sub-account performance for their return. The downline of the gold coin is also seen as a signal for the platform to reduce cash subsidies, to upgrade, to retain core users and to realize efficiency in advertising.
A windfall, because of the imposition of AI, the threshold is as low as it is accessible to all, is no longer a windfall, but a transfer of power. Huang Ho Nam predicted that, after the end of 2026, China would no longer have an AI comic agency and no longer had the logic of “organism”, and that the future would be full of OPCs.
When OPC becomes an increasingly common form of organization, the industrial structure of content production also moves from “institutional centres” to “individual centres”. The proceeds belong to those who run fast, to those who are lucky, and to those who sell shovels and pay for the road.
When the "genius" is no longer valuable, the "classic" is hard to be born, the "explosive" is becoming more and more like a probabilistic game, and the participants in the content industry may need to rethink: What are they producing for?
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.