Start with a number that reframes the entire video business.
As of July 2026, the microdrama app Hongguo had 168 million daily active users, up 107% year on year. iQiyi, Youku, Tencent Video and Mango TV combined had 155 million.
A vertical short-drama app had more daily users than the four major long-form platforms put together — with an average of 125 minutes per user per day.
Key takeaways
- Market size: China's microdrama industry reached roughly 120 billion yuan last year, against about 60 billion yuan for the entire Chinese theatrical box office. The "lowbrow" format is twice the size of cinema.
- The cost break: after ByteDance released Seedance 2.0 in February 2026, an AI-generated microdrama cost about one-tenth of a live-action one, and production time fell from roughly two months to about two weeks (reported by the Financial Times).
- Volume explosion: about 128,000 microdramas went online in China in Q1 2026, of which more than 95% were AI-generated. A year earlier, the whole industry produced around 3,000 a month.
- Team throughput: producer Guo Guo reports a ten-person team can now ship one 100-minute vertical drama per week, or one 45-minute episode of higher-quality long-form per week.
- Speed record: Hunan TV's fully AI-generated 30-episode series Journey to the West: Aftermath went from greenlight to broadcast in three months.
What "95%" actually means
It is worth pausing on that figure. It does not mean AI assists production. It means that among newly released short dramas, human-shot productions are now the minority.
The jobs that disappear are the obvious ones: on an all-AI project, art direction, set construction, camera crews, lighting and extras are largely not needed. Producer Liu Zhanhao describes the cost decline as "exponential."
But a new role has appeared on the other side of the ledger. The industry calls it the "card puller" — half prompt engineer, half visual editor. The job consists of feeding reference images and prompts into a model, adjusting parameters, and rerunning until a usable frame comes out. "Card pulling" is drawn from gacha games: you keep pulling until you get one worth keeping.
Where it still breaks
The people actually doing this work are the clearest about the limits. Liu Zhanhao names three unsolved problems: character and scene consistency, parsing of complex scripts, and video length limits.
Guo Guo adds the more important one: models still struggle with scenes of high emotional intensity.
That draws a clean boundary around current capability:
- Works: strong plot, fast pace, visual spectacle, genre content.
- Does not work: nuanced performance, emotional layering, long character arcs.
AI short drama's advantage is speed and price, not quality. The 95% penetration rate is partly because the content only has to be good enough — you swipe past after three minutes. The moment audiences start choosing carefully, that ratio will fall back.
So is there still a living in it?
Yes, but the entry point moved.
Do not compete on "I can use the tools." That skill saturates within months. What is scarce is aesthetic judgement — knowing which frame to keep and which to reroll. That is precisely the card puller's value, and it is where traditional film people have a stock advantage.
Move upstream. Script, storyboarding, pacing, character design — these are AI's weakest areas today, and roughly seventy percent of whether a short drama breaks out is story structure rather than imagery.
Treat 95% as a window, not a steady state. Supply is exploding and platforms are short of content, which makes this the easiest moment to enter. When platforms start selecting for quality, the bar rises quickly.
AI dismantled the barrier to making a film. It did not dismantle the barrier to making one worth watching.
Figures as reported by Shenzhen HuiDongChuangXiang, the Financial Times and Chinese industry media, 2026.