Cover Story: How AI Took Over China’s Micro-Drama Industry in 90 Days

1. The town of Hengdian, a major film hub in Zhejiang province, was previously booming with live-action short drama productions, with actors like Cheng Qiao earning 800 to 1,500 yuan daily and over 10,000 yuan monthly [para. 1][para. 2]. However, after the Lunar New Year in February 2026, work disappeared as AI tools, particularly ByteDance's Seedance 2.0, began transforming the industry's economics [para. 3][para. 4][para. 5][para. 6]. Productions that once required actors, crews, and days of shooting could now be generated by feeding prompts into video models [para. 6]. Additionally, ByteDance's distribution platform Hongguo Short Drama cut subsidies for live-action productions, leading some small companies to dissolve or switch to AI [para. 7]. As a result, many actors like Cheng are unemployed, cutting their rates to compete for scarce roles, and some crews pay below background actor rates [para. 8].

2. ByteDance now dominates China's short-drama industry with a closed-loop system covering scripts, AI models, cloud services, literature platforms, distribution, and advertising [para. 9][para. 10]. This ecosystem includes Fanqie Novel for copyrighted stories, Doubao and Jimeng AI tools, Volcano Engine cloud services, Hongguo and Douyin for distribution, and Ocean Engine for advertising [para. 46][para. 47][para. 48]. In 2024, ByteDance bet on model-as-a-service (MaaS), and by March 2026, daily token usage for Doubao's large model doubled to over 120 trillion [para. 49][para. 50]. Volcano Engine led enterprise MaaS call volume in 2025 with 49.5% market share, far ahead of Alibaba Cloud's 28% [para. 51]. A policy researcher noted that downstream distribution via Hongguo and upstream model access via Seedance 2.0 channel a large share of value chain money to ByteDance, disrupting the ecosystem that previously supported actors and screenwriters [para. 52][para. 53][para. 54].

3. The explosion of AI short dramas stems from advances in video generation models like ByteDance's Seedance 2.0, released in February 2026, which exceeded market expectations in realism and quickly became mainstream [para. 16][para. 17]. Demand was so high that generating a single shot could take over an hour, and prompt engineers worked overnight to avoid peak demand [para. 18]. Short dramas remain volume-driven: in Q1 2026, about 128,000 micro dramas were released nationwide, nearly four times the total of the previous year, with 122,000 (over 95%) being AI-made [para. 19][para. 20]. A Xi'an executive estimated a 70-80% drop in live-action production monthly, but simulated-live-action capacity surged, with nearly 50,000 new dramas online in April [para. 21]. The audience remains over 40 years old in lower-tier cities, caring mainly for emotional story impact, but AI shows attract younger viewers due to genre diversity [para. 22][para. 23].

4. Cost reductions are dramatic: a Beijing producer said larger companies in Hangzhou can make a 100-minute AI short drama every 3-7 days for under 20,000 yuan, with college graduates earning 5,000 yuan monthly pretax [para. 24][para. 25]. Even mid- and senior-level staff earning 40,000 yuan are under pressure, as AI replaces years of content judgment with data [para. 26]. Team sizes are shrinking: three months ago, an animated drama team had six members; now one-person companies (OPCs) handle the entire process [para. 27]. Animated show director Zhang Lutian reported that video generation accounts for over 95% of AI short drama costs [para. 28]. Contract production companies have pushed prices from 800-1,000 yuan per minute to 200 yuan per minute, driving prices close to computing power costs [para. 29][para. 30].

5. Seedance 2.0's long-term outlook is uncertain: a person familiar with ByteDance's model team said generative efficiency is high, but content quality does not differ much, and training costs are substantial [para. 35][para. 36]. The model's standardization leads to similar outputs if prompts are insufficiently specific [para. 37]. The Xi'an executive noted breakout hits are rare, and genre replacement cycles are accelerating, requiring constant script adjustments [para. 38].

6. ByteDance's funding policy drove Hongguo Short Drama to over 100 million daily users in January 2026 [para. 39]. In 2025, the company invested 2.5 billion yuan, offering 200,000-350,000 yuan per title to production contractors, covering costs and leaving surplus, with revenue sharing [para. 40]. However, an internal review concluded incentive policies were too broad and paid user acquisition put financial pressure on the company [para. 41]. Hongguo became profitable in 2025 ahead of target, so it set aggressive 2026 profit goals [para. 42]. In late February 2026, Hongguo canceled guaranteed partnerships with mid-tier live-action contractors amid the AI shift [para. 43]. A live-action short drama required a week of prep, 3-5 days of shooting, a crew of nearly 40, and budgets of 300,000-400,000 yuan in 2025 [para. 44]. Now an AI drama can be produced with fewer than 10 people for under 100,000 yuan, some under 20,000 yuan [para. 45]. After ByteDance's cuts, a Zhengzhou-based producer halted live-action projects, gave up deposits, switched to AI, lost nearly half its staff, and had the producer's salary drop below 50% [para. 55][para. 56][para. 57]. The producer noted AI results were better and cost lower, so the switch made business sense [para. 58]. Hongguo later introduced new policies to support live-action dramas, with a slight pickup since May, but only for premium or low-budget projects (~200,000 yuan) [para. 59][para. 60]. Industry consensus is that reversal won't restore live-action volumes to 2025 levels if downstream buyers no longer pay [para. 61].