--- title: "One-click microdramas: how China's AI video models turned short drama into an export" date: 2026-09-24 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-aigc-ai-microdrama-overseas language: en --- # One-click microdramas: how China's AI video models turned short drama into an export > At a Shanghai conference on 4 September 2026, China's AI microdrama (短剧) overseas push got a policy-and-ecosystem launchpad, backed by real money — US$1.27 billion in overseas in-app revenue in the first half and a market now forecast above US$6 billion. The engine is generative video, led by ByteDance's Seedance 2.5. A Chinese short drama now lands in a foreign viewer's feed within hours of being written — sometimes with the script, the faces, the voices and the subtitles all generated by models. The "Chinese story, made by AI, watched worldwide" pipeline is no longer a pitch. It is a supply chain. On 4 September 2026, Shanghai's Lingang district hosted the 2026 AI+Microdrama Overseas Industry Promotion Conference. The event itself is a sign of where the money and policy attention are pointing: from domestic short-video classes to a state-backed push to export AI-made microdrama (短剧). ## The Shanghai launchpad The conference was not just speeches. Lingang opened an AI+Microdrama overseas service station with linked sites in Riyadh, Singapore, the UK and Switzerland, offering compliance navigation, localisation and co-creation support. Local officials pitched a "policy gift pack" of fiscal support, talent guarantees and easier cross-border data flows for digital-culture exporters. The hook is the revenue. According to figures cited at the conference and reported by China Daily, overseas in-app purchase revenue from Chinese microdrama (短剧) reached US$1.27 billion in the first half of 2026. That is the number that turns a content trend into an industrial strategy. ## The machine behind the flood None of this scales without the generative models. The clearest technical step this quarter is ByteDance's Seedance 2.5, a video-generation large model (大模型) that can produce a continuous 30-second scene in a single pass — up from the 5-to-15-second clips that defined AI video a year ago — while ingesting up to 50 multimodal reference assets (images, video, audio) so characters and settings stay consistent. ByteDance ships it inside Jimeng (即梦) in China, Dreamina and CapCut internationally, and through its Volcano Engine (火山引擎) enterprise API. Reports describe a "Drama Creation" mode aimed at one-click short-drama output: feed the model a script and reference images, get a coherent scene back. Kuaishou's (快手) Kling 3 and Alibaba's Wan 3.0 push the same direction — longer, steadier, more controllable clips. The practical effect is brutal for traditional production. What used to take a shoot, a cast, locations and post-production now starts as a prompt. Industry reports describe a three-person team funded with roughly 100,000 yuan producing a dozen seasons of an AI microdrama (短剧) with overseas returns reported in the hundreds of thousands of dollars — an anecdote, not a typical outcome, but a fair illustration of how far the cost floor has dropped. ## The money is already moving The market data is the part to take seriously. DataEye (剧查查) estimated overseas microdrama (短剧) monthly active users at 258 million and raised its 2026 overseas market forecast from US$5 billion to above US$6 billion. Its AI-drama-and-manga slice alone is put at more than US$4 billion, up roughly 390% year on year. Public companies are booking the revenue. Kunlun Tech (昆仑万维) reported first-half 2026 AI microdrama platform revenue of 1.534 billion yuan (about US$216 million / HK$1.66 billion), up 163% year on year, with its DramaWave platform saying over 90% of July's new titles were AI-generated. Chinese Online's (中文在线) FlareFlow passed 52 million registered users and nearly 7,000 titles, with AI shows above 10% of the catalogue. The caveat is the one every gold rush shares: most of the money is concentrated. Analysts note overseas short-drama downloads lean 70–80% on paid user acquisition, and roughly 90% of revenue sits in the US market. A microdrama (短剧) can be generated in one click; finding paying viewers still costs real dollars. ## What readers can do now - **If you make content:** treat the model as a pre-visualisation and localisation engine, not a finished product. The winners pair AI throughput with a human who owns taste and cultural fit. - **If you localise for export:** the leverage is in multi-language dubbing, lip-sync and subtitle automation — that is where AI removes the old "miss the trend window" problem. - **If you watch the sector:** track overseas paying-user acquisition cost, not just view counts. A viral AI microdrama (短剧) with no retention is a science project, not a business. ## Honest limitations Conference, policy and ecosystem details (4 September 2026 Lingang event, overseas service stations, US$1.27 billion H1 overseas in-app revenue) are reported by China Daily and Chinese industry media; the revenue figure is an industry estimate, not an audited aggregate. Seedance 2.5 capabilities (30-second single-pass generation, up to 50 reference inputs, native audio, multilingual) are from ByteDance and Chinese state-media and trade reporting (CNR, provincial newspapers) and third-party analyses; the "one-click full drama" framing is ByteDance/Volcano Engine positioning reported by secondary outlets, not an independent benchmark. Market-size and user figures (258 million MAU, US$6 billion 2026 forecast, US$4 billion AI slice, +390%) are DataEye estimates cited across Chinese media. Company revenue (Kunlun 1.534 billion yuan, +163%; FlareFlow 52 million users) are disclosed company figures reported by securities and tech media. The three-person-team example is a media-reported anecdote, not a typical result. Currency conversions use roughly 1 USD ≈ 7.1 RMB and 1 RMB ≈ 1.085 HKD. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-aigc-ai-microdrama-overseas Free to quote with attribution and a link to the original.