Quiet on Set: How AI Transformed China's Microdrama Scene

A few months ago, the studios in Hengdian, the film town people call China's Hollywood, were packed. Crews shot fight scenes next to palace-style sets. Extras waited around in full costume. Then, within a few months, that work started to dry up.

It wasn't because people got bored of the shows. It was because a new tool made it possible to skip the sets altogether.

That tool is AI video generation, and China's microdrama industry has been hit by it hardest. These are short, vertical shows built for phones, and what happened to them raises a bigger question: once video gets cheap and fast to make, what still makes it good?

China's Microdrama Boom, Briefly

Microdramas are called duanju in China. Short videos are made for phones, and they are filmed in a vertical frame. They really took off during the pandemic. Back then, apps like Douyin and Kuaishou were growing fast, while most normal film shoots had mostly stopped.

Each episode runs a minute or two. But a full series can run up to 100 episodes. The plots are soapy and full of twists, and each one ends on a cliffhanger that keeps people tapping "next."

The format grew fast. China's microdrama market more than doubled in 2025, reaching over 100 billion yuan, or roughly $14.4 billion. Almost 700 million people watched microdramas that year. For a while, the boom created real jobs too, for actors and crew who'd never worked in traditional film.

How AI Moved From Novelty to Standard Practice

The shift happened fast, and new tools made it possible. ByteDance’s Seedance 2.0, Kuaishou’s Kling, and Shengshu’s Vidu have gotten better. The scenes can now look like they were shot by a real crew. In early 2026, about 128,000 fresh microdramas were released in just one quarter. More than 95 percent were made with AI. The shift did not take long. In roughly one year, team workflows changed a lot.

This also clarifies what people mean when they say “AI-generated”. Some folks use that phrase for more than one setup. And those setups are not always the same.

• Fully AI-generated microdramas build entire scenes and characters from text prompts. No cameras or sets needed.

• AI-assisted production keeps real actors and directors in place. It just leans on AI for concept art or effects.

• Traditional live-action still uses real performers and real sets. AI shows up later, mostly in editing.
  

Most coverage focuses on the first kind. One production firm says it can now finish a full 120-minute project with a five-person crew in about a week. The cost: roughly $2,200, a fraction of what a low-end live-action shoot runs. Staff called "card-pullers" write and rewrite prompts until a clip matches the scene. A small creative team handles the rest, work one founder compared to casting and location scouting, just done on a screen.

Why the Economics Push So Hard Toward AI

Microdramas were never a prestige format. They ran on volume, with companies churning out titles and betting that one would go viral. That model makes AI's math hard to ignore.

• Production costs have dropped to around a fifth of traditional shoots.

• Turnaround has shrunk from roughly three months to about one.

• Local governments have started chipping in too, offering extra funding worth up to two million yuan per approved drama.

• The overseas market for AI-made dramas is expected to jump from $100 million in 2025 to $650 million this year, according to DataEye.
  

When a format already rewards speed and volume, and AI cuts costs this much, switching becomes an easy call.

What AI Genuinely Does Well Here

Give it credit where it's due. AI has made this corner of production faster, cheaper, and easier to experiment with. Concept art that once needed a dedicated artist now takes minutes. AI has already replaced most of the artists who used to shape a project's early look, according to one crew member in Shanghai.

Script "optimization," compressing a novel into a tight outline, now takes a fraction of the time a writer needs. And because generation is cheap, teams can try a dozen versions of a scene before committing to one. That was never realistic when every take meant paying a crew for another hour.

What AI Still Can't Do

Here's where this gets more interesting than a simple "AI replaces people" headline. The people living through this shift don't talk about it that way. AI can generate a hundred versions of a fight scene, but it can't tell you which one actually serves the story, or whether the story is worth telling at all.

The Details a Prompt Can't Fake

A real interview brings out details an AI clip can't fake. Think of a hesitation before a hard answer, or a small detail only someone who lived through the moment would think to mention. Genuine B-roll works the same way. It shows what a business actually does, not a generic stand-in for it.

Knowing What the Video Is For

AI doesn't know who a video is for. It can spit out options, sure, but understanding a client's audience, and the goal behind the request, still takes a person. The same goes for knowing what to leave out. That's just as much a creative call as knowing what to show.

What Audiences Still Choose, and Who Owns a Face

Viewers seem to notice the difference too. Even with AI titles flooding platforms, "real people dramas" still rank in the top 10 on at least one major streaming app. One actor summed up the tension well: AI is formulaic, she said, and it can't express real emotion, so it can't replace a person. It might replace her job, though, in a crowded market.

There's also a messier problem. At least one well-known actor has already won court-ordered pay after an AI character was built to resemble them without permission. That's a question the industry still has to sort out.

What Professional Video Producers Can Take From This

The lesson here travels well beyond microdramas.

1. Cheaper production doesn't make better stories on its own. It just lowers the cost of testing whether a story works, which raises the value of knowing what's worth testing.

2. AI is good at tasks that were already mechanical. Generating concept variations is different from deciding what a video needs to accomplish. That decision still belongs to real pre-production, not a prompt.

3. Authenticity is becoming a differentiator. As AI footage gets harder to tell from the real thing, planned, on-location footage matters more, not less.

4. The human moment still has to be earned. No tool decides who belongs on camera, or what makes someone feel believable once they're there.

Producing More Video Versus Producing Better Video

China's microdrama industry answered a narrow question fast: can AI make cheap, throwaway content even cheaper? Clearly, yes. The bigger question is taking longer to answer. Does cheaper production actually make anything worth watching?

The technology can generate footage. It still can't decide what that footage is for. That's the gap professional video work lives in now. As the tools for making video get faster and cheaper everywhere, the harder work shifts earlier, into figuring out what story actually deserves to be told.

Where AI Fits Into Video Production With 24Frames

AI has a real place in professional video work. It speeds up parts of the process that used to eat time and budget. But it works best when it serves a clear creative goal. At 24Frames, that means starting with the story, the audience, and what a video needs to do, before deciding which tools belong in our process. That's still where solid brand work begins.

Frequently Asked Questions

What are Chinese microdramas? 

They're short, vertical videos made for phones. Episodes usually run 60 to 90 seconds, building one long story across dozens of parts.

How is AI being used in Chinese microdramas? 

AI tools generate footage, characters, and full scenes from text prompts. That cuts production time from months to weeks.

Why are microdramas so popular in China? 

They fit how people already watch video, in short bursts on a phone. Cliffhangers keep people coming back for more.

Can AI make an entire movie or video? 

Yes, at least for microdramas. AI can generate a full episode at a time, and most new releases in the format are now fully AI-made.

Will AI replace actors and filmmakers? 

Reporting so far shows AI replacing specific jobs and shrinking budgets. It hasn't replaced the creative decisions behind a production.

What can businesses learn from AI-powered video production? 

Cheaper production doesn't automatically mean better content. As video gets cheaper to make, knowing what story to tell becomes what actually sets it apart.