AI video production is shifting toward a "screenwriter-centric" model, says Wu Gaoming of Jingying Tech.

Wu Gaoming, co-founder of Jingying Technology, believes that AI can replace directors, actors, and post-production, but cannot replace screenwriters. In the first quarter of this year, China launched 1.28 million new micro-short videos, over 95% of which were AI-generated, with 718 million monthly active users. Jingying Technology’s Reel.AI series “The Billionaire’s Return” topped overseas short-form video charts in November 2025, proving that AI-generated shorts can now compete with live-action content. The source of high-quality creativity has not been structured and made available online; these implicit knowledge models cannot be learned by AI. As backend production processes are taken over by AI, the industry’s focus is shifting toward the frontend—what story to tell and to whom has become the most valuable环节, and AI video is moving toward a “screenwriter-centric” model.

Article author and source: Wall Street Journal

AI has replaced directors, actors, and post-production, but not screenwriters.

This is the assessment by Wu Gaoming, co-founder of Jingying Technology.

In the first quarter of this year, China launched 128,000 new short-form dramas, over 95% of which were AI-generated. Monthly active users reached 718 million, with an average daily viewing time of 129 minutes per person—surpassing long-form video for the first time. Global in-app purchase revenue from short-form drama apps reached $2.98 billion last year, more than doubling year-over-year.

On the content creation side, the entire production chain is being taken over by AI, but Wu Gaoming believes that screenwriting is the final and most difficult link to be replaced.

Because the sources of creativity, intuitive understanding of human emotions, and experiences of the world... these things are not structured and placed on the internet, so the model cannot learn them.

Jingying Technology was founded in 2021 and is among the first companies globally to successfully implement a paid AI short-form drama business model. Its platform, Reel.AI, targets overseas markets; according to public reports, in November 2025, the platform’s production “The Billionaire's Return” topped the overseas short-form drama charts, outperforming live-action dramas.

For the first time, an AI short drama has accomplished this.

But Wu Gaoming does not define Jingying Technology as an "AI short drama company." The backend processes—video generation, voiceover, and editing—are handled by AI, while storytelling and aesthetics are left to creators. The platform provides computing power, tools, and distribution channels; creators only need to focus on one thing: telling a great story.

In June this year, Jingying Technology completed its Series A and Series A+ funding rounds, raising tens of millions of dollars. Investors include Wang Huiwen’s family office, Lollapalooza Capital, Ant Group, and Yin Yu, former vice president of Tencent.

The next phase of competition may be decided on the side of creators.

The order in which AI replaces content production各个环节 is from back to front.

Models are trained on data; the later the stage, the more data available, and the faster it is learned. With vast amounts of video material in later stages, the earliest elements to be replicated will fall first; actors' performances and directors' stylistic techniques are also discernible patterns.

Wu Gaoming has a saying: “Film is inherently an ‘art of deception’—actors use performance to immerse audiences in their roles, and directors use visual language to evoke emotion. AI doesn’t differ fundamentally from humans in these ‘deceptive’ processes; if it’s convincing enough, audiences will accept it.” User data from Reel.AI supports this: users neither avoid content simply because it’s AI-generated, nor champion it solely because it is. “In the end, as long as it satisfies user preferences, they don’t care about the underlying production method,” Wu Gaoming said.

However, in the short term, AI struggles to replace screenwriters.

“Many sources of high-quality ideas have not been structured or digitized on the internet,” said Wu Gaoming. “Why do people find this story compelling? Why do they feel compelled to create a particular story? These are human tacit knowledge.”

Models naturally tend toward a central distribution; after extensive training, outputs converge toward the most common appearance. This is how “AI faces” emerge—models pull all faces toward the average.

But the content industry does not reward averages.

It rewards sharp, off-center content infused with the creator’s unique experience. On Jingying Technology’s platform, a short drama set in an Antarctic research station emerged—a story about an American scientist and a Russian scientist stranded in a frozen wasteland. This kind of subject matter would never appear on traditional film and television schedules, and top-tier producers wouldn’t fund it. Yet it resonated with the authentic needs of a group of users.

What captures this need is not the model, but the screenwriter’s intuition for what is worth telling.

As AI gradually takes over backend production processes, the focus of the entire content industry is shifting toward the frontend—deciding what stories to tell, to whom, and how to tell them is becoming the most valuable环节. Wu Gaoming believes this is the logic behind AI video moving toward a “writer-centric” model. After the director-centric model is dissolved by AI, the weight of creative judgment increases rather than decreases.

For Jingying Technology, once the technical barriers were lowered, the real challenge became finding people.

The team previously ran one of China’s top three free novel platforms, and that experience left them with a methodology: don’t wait until content is fully completed to test it—test a small portion first.

In 2024, almost no one believed in AI short films. Wu Gaoming was looking for creators who had a background in film and television and genuinely believed AI could produce something great—such people were extremely rare. He searched online for creators with film and television experience and sent out questionnaires en masse. Hidden within the questionnaire was a key question: Evaluate an AI-generated short film already producible by Jingying Technology—its visuals were crude and full of flaws.

What he’s looking for is the reaction after pointing out the flaws—whether you’ll find ways to work around AI’s limitations, and whether you truly believe this can ultimately succeed.

Over 1,000 people completed the survey, and ultimately, 10 remained.

One of them was Choi. He later created the platform’s first AI short drama to achieve commercial success overseas.

Jingying Technology does not produce any in-house content; creators are almost entirely external. The revenue-sharing model is a guaranteed minimum plus profit sharing: creators receive the guaranteed minimum after their outline is approved, again upon completion of the script, and additional profit shares once the content is live and generates revenue. Creators earn at every stage, with the platform bearing all risks.

After implementing the system, Wu Gaoming found two types of people particularly well-suited to it: those with experience in the traditional film and television industry but who never made it to the top, and those who studied film production abroad but found themselves unable to break into China’s domestic industry upon returning. For individuals who were overlooked in the old system, AI provided a new opportunity to re-enter the field.

Once the right people are in place, the efficiency gap quickly widens. Traditional live-action short dramas cost between $150,000 and $300,000 per episode and require approval from top management, with each validation cycle taking at least three months. Jingying Technology, however, only produces the first 8 to 10 episodes for market testing, monitoring conversion rates, user feedback, and payment signals—once the data is in, they know whether to continue. “Others need at least three months for one round of testing,” Wu Gaoming said, “but we might only need a week.”

Here, AI approximates a flexible supply chain—front-end experimentation at extremely low cost, with backend production scaling up instantly upon success. Jingying Technology also has an internal "content evaluator" that can predict market response before the creative content reaches users.

The company now generates millions of dollars in monthly revenue. In January 2025, "Five Brothers" became the world's first AI short drama to compete on the same chart as live-action short dramas; in November of the same year, "The Billionaire's Return" reached the number one spot.

In the early days, Jingying Technology made a misjudgment due to data noise in the payment process.

A group of users who appeared to be paying were actually engaging in fraud. The team attributed the issue to content quality and conducted a lengthy review but could not identify the cause. Eventually, they discovered the problem lay with the payment signal.

Jingying Technology chose Stripe. Initially, it accounted for only about 30% of total volume, with the company simultaneously integrating other channels and managing routing internally. However, as data accumulated, it became clear: users on the Stripe channel had significantly higher renewal rates, with a long-term ROI difference of approximately 20%.

“This routing result was not manipulated manually,” said Wu Gaoming. The percentage of Stripe’s transaction volume naturally rose to nearly 70%.

The underlying logic is that Stripe’s fraud detection system, Radar, filters out high-risk behaviors during the registration phase. Melina Lee, General Manager of Enterprise Customers for Stripe in Greater China, who has collaborated with Jingying Technology for many years, shared that for AI applications, once users begin using the platform, it triggers actual token consumption for the business—costing significant money. Radar can proactively detect potential token theft, malicious free-trial sign-ups, and other abusive behaviors, enabling businesses to anticipate and mitigate risks in advance.

At the payment level, Smart Retries automatically retries failed debit attempts at the optimal time, recovering many subscribers who would otherwise have churned due to card issues rather than intentional cancellation. Liu Sicheng, Head of Enterprise Sales for Stripe in Greater China, told Wall Street Journal that Stripe maximizes the likelihood of users completing subsequent renewals. For example, during the first payment, card balances are typically sufficient and outcomes are consistent; however, during the second payment, the card may have expired or insufficient funds. Each service provider handles this “trial-and-error” and retry process differently. Because Stripe processes vast volumes of data and serves a large, diverse consumer base, it can more accurately identify the optimal timing for debit attempts, ensuring successful renewals. This translates to higher customer value for merchants. Smart Retries is designed to maximize payment success rates throughout the renewal process. “This capability also relies on Stripe’s large payment model, and our ‘training corpus’ is the payment data itself.”

For Jingying Technology, when the payment data is clean, content decisions become accurate. Whether to produce a drama or allocate resources depends entirely on user signals. If the signals are corrupted, the decisions will be wrong.

In 2026, Wang Minjie, former Chief Application Scientist at AWS, joined Jingying Technology as Chief Scientist. Founder Zhu Jiang believes that the Coding Agent has already proven that agents can deeply transform an industry. “The content industry is next.”

The AI content industry is still in its very early stages, and it's far too soon to conclude what the eventual leading companies will look like.

Jingying Technology’s exploration may also be answering a more fundamental question for the content industry: once AI takes over the execution of content creation, will the real competition ultimately come down to who can tell a better story?