Technology Insight

AI-Generated Video in China Is Replacing Actors—and Rewriting the Economics of Digital Media

ByteDance’s Seedance models and a new generation of digital performers are collapsing production times and costs. The result may be a margin opportunity for platforms—and a labor, copyright, and content-quality crisis for the creator economy.

Bottom line: Artificial intelligence is turning video production from a labor-intensive creative business into a software-driven, high-volume operation. Investors should watch the platforms and infrastructure providers that capture the savings—but also the legal and regulatory liabilities that come with digital replicas.

A new phase of artificial intelligence has arrived in China’s vast online-content economy. Generative video systems can now create short dramas, animated series, advertisements, and livestreaming hosts with a speed and level of visual consistency that were out of reach only a short time ago.

The scale of adoption is striking. According to figures reported by the Financial Times, approximately 128,000 short dramas were released in China during the first quarter of 2026—more than three times the number released during all of 2025—and 95% were generated with artificial intelligence. In May, 89 of the 100 highest-ranked animated dramas on Douyin, ByteDance’s domestic counterpart to TikTok, were AI productions.

This is not simply a new special-effects tool. AI-generated video in China is changing the unit economics of content creation, shifting bargaining power toward platforms and model owners while placing actors, influencers, editors, animators, and production crews under immediate pressure.

Seedance Changes the Production Equation

ByteDance officially launched Seedance 2.0 in February 2026. The model combines text, image, audio, and video inputs within a unified generation system, allowing users to control performance, lighting, camera movement, sound, and visual references. In July, ByteDance introduced Seedance 2.5, extending single-pass generation to 30 seconds and improving the model’s ability to organize multiple shots into a coherent story.

Those capabilities attack several costly stages of conventional production at once. A creator can move from concept to storyboard, casting, performance, cinematography, voice, sound design, and editing without assembling a traditional crew for every step.

Professor Shen Yang of Tsinghua University estimated that a three-to-five-minute drama that required five people and three months in 2024 can now be produced by one employee in one or two days. Reported production prices of Rmb600–Rmb800 per minute are roughly 10% of the cost of using human performers and conventional workflows.

Production factor	Traditional workflow	AI-driven workflow	Economic effect
Labor	Actors, camera, lighting, sound, editing, and production staff	One creator can direct multiple software tools	Lower costs; fewer entry-level and freelance jobs
Time	Weeks or months	Hours or days	More experiments and faster response to trends
Output	Constrained by schedules, locations, and reshoots	Scalable, localized, and continuously revised	Content supply can expand dramatically
Commercial testing	Limited number of expensive versions	Many low-cost variants	Rapid optimization of ads, stories, and sales scripts

The Human Cost Is Larger Than the Acting Industry

China’s short-drama business directly employed about 690,000 people in 2025, according to research from Peking University’s National School of Development. That estimate includes roughly 450,000 production workers, 32,000 lead and supporting actors, and 207,000 extras. Including indirect work across related industries, the employment footprint may exceed 2 million.

Livestreaming is larger still. Approximately 15 million people identified it as their primary profession in a 2024 industry report. That workforce includes hosts, assistants, product specialists, agencies, set designers, makeup artists, camera operators, moderators, and sales personnel.

The disruption is already visible. Beijing-based actor and producer Greg Wollner said his schedule fell from three productions a week to almost nothing after Seedance 2.0 appeared. Actor Todd Kuhns described a broader feeling of frustration and resignation among performers who see their livelihoods being removed by systems capable of imitating human work.

The near-term danger is not that every creative job disappears simultaneously. It is that fewer people are needed for each unit of content, while a much smaller number of AI-literate producers can generate a much larger volume. That transition can create more output and higher platform revenue without creating comparable employment.

Digital Influencers Turn Attention Into a Software Asset

The most commercially important development may be occurring in livestream commerce. Human hosts face obvious constraints: they need rest, their schedules are limited, and their performances can vary. Digital hosts can operate continuously, use multiple languages, update product information instantly, and run many simultaneous streams.

Baidu demonstrated the potential with digital versions of entrepreneur and influencer Luo Yonghao and his co-host. Their six-plus-hour livestream attracted more than 13 million views and generated more than Rmb55 million—about $7.7 million—in gross merchandise value. In several categories, the avatars reportedly outsold Luo’s earlier human-hosted debut.

That result turns a creator’s identity into a scalable commercial asset. A licensed digital double could appear in numerous stores, languages, or markets at the same time. The platform can test scripts and offers continuously, while the human personality becomes intellectual property rather than the person who must perform every session.

Why This Matters to Investors

The central investment question is not whether AI-generated video in China will produce more content. The evidence already points strongly in that direction. The harder question is who captures the economic value when the marginal cost of content approaches the marginal cost of computing.

Potential Beneficiaries

Model and platform owners: ByteDance can combine generation tools with Douyin’s distribution, advertising data, and recommendation engine.
Search and e-commerce platforms: Baidu’s digital-host technology can lower merchant acquisition and livestreaming costs.
Cloud and computing providers: Expanding video generation increases demand for inference, storage, networking, and specialized processors.
Merchants and advertisers: Low-cost creative variation can improve localization, product testing, and conversion optimization.
Rights owners with trusted brands: Actors and creators who control valuable likeness rights may license digital doubles across many campaigns.

ByteDance is privately held, so public-market exposure is indirect. Traders watching the theme may focus on Baidu, Tencent, Kuaishou, Bilibili, Alibaba, JD.com, Chinese cloud providers, and semiconductor or data-center companies supplying the computational layer. The presence of a theme, however, does not make every company associated with it an attractive trade. Investors still need evidence of revenue growth, improved margins, disciplined capital spending, and defensible intellectual property.

The volume paradox

When content becomes extremely cheap, supply can rise faster than human attention. That can reduce the value of any individual video and force platforms to spend more on recommendation systems, moderation, and discovery. Lower production costs may improve gross margins for creators while simultaneously making it harder for them to earn money because competition becomes nearly unlimited.

Platforms are better positioned because they control distribution and can monetize the aggregate increase in engagement and advertising inventory. This creates a familiar digital-economy pattern: the tools democratize production, but the gatekeeper may capture the most durable value.

Copyright and Likeness Risk Could Change the Winners

The efficiency story comes with substantial legal risk. Hollywood groups and actors’ representatives have criticized Seedance 2.0 over unauthorized depictions of well-known performers and copyrighted characters. Disney and other studios challenged ByteDance, and the company said it would strengthen safeguards against improper use of intellectual property.

These disputes point to several unresolved questions:

Was copyrighted material lawfully used to train the model?
Who owns a generated performance based on a real person’s face, voice, or mannerisms?
Must viewers be told that a host, testimonial, or performer is artificial?
How should actors and influencers be compensated when licensed replicas work continuously?
Who is liable when generated content makes a false claim, violates advertising rules, or damages a person’s reputation?

For investors, safeguards are not merely an ethical feature. They may become a competitive moat. Platforms that can document consent, provenance, training rights, and compensation could be more attractive to major advertisers and global media partners than systems built for speed without reliable rights management.

Trading the AI-Video Theme

AI-generated video in China is a powerful structural narrative, but the stocks connected to it can remain highly sensitive to Chinese regulation, US–China trade policy, export controls, consumer spending, and company-specific earnings. Traders should separate long-term technological potential from the conditions required for a high-quality short-term setup.

TraderInsight Market Checklist

Identify the catalyst. Look for a product launch, adoption metric, earnings commentary, partnership, or regulatory decision.
Find the economic proof. Revenue, merchant adoption, engagement, conversion, and margin expansion matter more than impressive demonstration videos.
Check related names. Confirmation across platforms, cloud providers, and semiconductor suppliers can distinguish a sector move from a single-stock reaction.
Map the technical levels. Use premarket support, resistance, volume-by-price, and order flow to define the point where the market confirms the thesis.
Respect headline risk. Copyright restrictions, model suspensions, and new disclosure rules can reverse sentiment quickly.

A bullish scenario would combine strong adoption data with credible monetization and limited regulatory resistance. A bearish scenario would emerge if content volume grows but revenue per creator or advertiser deteriorates, computing costs remain high, or legal restrictions block commercial deployment. The most volatile scenario may be the middle one: rapid adoption alongside unresolved intellectual-property disputes.

The Broader Economic Signal

China’s short-drama and livestreaming markets offer an unusually clear preview of how generative AI may affect service-sector work elsewhere. The technology does not need to reproduce the very best actor, director, or influencer to cause disruption. It needs only to be good enough for high-volume commercial content while being faster, cheaper, and endlessly scalable.

That threshold appears to have been crossed in important parts of the Chinese market. The next phase will determine whether displaced workers move into higher-value creative roles, whether new categories of work emerge, and whether regulators require the productivity gains to be shared with the people whose faces, voices, and performances made digital replicas valuable.

The Trading Takeaway

AI-generated video in China demonstrates how quickly an AI capability can become an economic substitution event. Production has not merely become easier; the relationship among labor, time, cost, and output has been rewritten.

The strongest long-term companies may not be those that generate the most visually impressive clips. They may be the businesses that combine capable models with distribution, proprietary data, commercial customers, rights management, and an economically sustainable computing stack.

For traders, the opportunity lies in recognizing when that structural story becomes a measurable catalyst—and then waiting for price, volume, and order flow to confirm that the market agrees.

TraderInsight principle: Technology creates the narrative. Adoption, monetization, and price action determine whether the narrative becomes a trade.

Sources

Disclaimer: This material is provided for educational and informational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security. Trading involves risk, including the possible loss of principal. Company, regulatory, and market conditions can change rapidly.

AI-Generated Video In China