China Put a Huangmei Opera on a Vertical Screen, and the Singing Survived

"Da Zhu Cao," or "Pulling Pigweed," is a Huangmei opera standard. On 15 September it arrived on Anhui Satellite TV's Haitu micro-drama channel as a short-form series, made with digital humans, motion capture and rendered environments. The broadcaster says it is the first AIGC Huangmei opera micro-drama in the country. It kept the Susong dialect and the caicha melody, the two things that make the tune recognisable to anyone from the region.

That last detail is the whole point. Plenty of AI video projects in China this year have used traditional culture as wallpaper, generating a courtyard, a robe and a lantern and calling it heritage. This one had to hold a musical form together inside a format built for phones.

Why a dying opera and a phone screen fit

Huangmei opera is not a museum piece in Anhui, but its audience was ageing and its venues were local. Huaining County's Huangmei opera troupe has livestreamed on Douyin since 2022. By 2026 its official account had 1.395 million followers and more than 600 livestreams with over 100 million cumulative views. Younger viewers went from around 10 percent of the audience to 45 percent. The troupe built a 5G studio and brought 4K capture and streaming gear into a county-level company, and it split classic arias and plot fragments into clips that fit a phone. The conversion runs in both directions. Some viewers hear an aria online first and then buy a ticket to a theatre, which turns a streaming account into a box office channel rather than a replacement for one.

The "Da Zhu Cao" production sits in that same shift. It uses a nested structure, with a modern storyline framing a period performance, and it renders the mountain hamlets and terraced fields of the original setting rather than filming them. Anhui Broadcasting's description is that AI restored the village scenery while the vocal style was preserved as it stands.

There is a version of this story where the technology replaces the performers. That is not what happened here, and the reason is practical rather than sentimental. Motion capture needs a body to capture. Digital humans need a voice to imitate. The parts of Huangmei opera that carry meaning, the dialect, the ornamentation, the timing of a comic beat, are exactly the parts that had to be sourced from people who can still do them.

The same month produced a whole shelf of them

"Da Zhu Cao" was not alone. It appears in the National Radio and Television Administration's second "micro-drama plus" recommendation list, one of 165 titles. "Sai Shang Yu Ge," an AI micro-drama about 30 years of Fujian-Ningxia cooperation, went live in Fuzhou on 6 September and turned a Gobi waste into a fishing ground on screen, setting the dry north-west beside the wet south-east. "Tao Hua Tan Ji" used a fully AIGC pipeline to walk through the 108 traditional steps of making Xuan paper, wrapped in a suspense plot. Others, including "Xiao Cheng Ke Ai" and "Jia Ou Tian Cheng (San Yue San)," pushed Guangxi mountain imagery and Zhuang brocade motifs onto TikTok and Melolo for audiences in Southeast Asia.

Heritage became the theme of the season rather than a single project. "Ye Feng Ye Ceng Rao Zhi Rou" built a story around dough figurine craft and a craftsman who uses livestreaming to restart a business. "Huan Gui," shot horizontally, took the suona tradition of western Liaoning and paired it with a "big screen leads small screen, IP merges with industry" distribution model. "Cha Xue Yuan Li Cang Zhe Bai Nian Zong Shi" used Liubao tea culture as its foundation and a university campus as its delivery.

The industry's own description of this is that "micro-drama plus travel" has become a standardised route for local cultural promotion. A county with a real story and a heritage asset can now produce something a provincial broadcaster will air, because the production cost that used to block it has collapsed.

The regulatory backdrop

The Administrative Measures for the Development of Micro-Dramas took effect on 1 September with eleven defined red lines. In the first eight months of the year, 430,000 micro-dramas went live, thirteen times the whole of last year, and AI-made titles were more than 90 percent of that volume.

The administration's public position is that live-action drama remains the main force for high-quality work and will be supported, while AI drama gets encouragement and limits at the same time. Its phrasing is that AI is a tool and people are the subject of creation. That sentence is why a provincial network could fund an opera project and expect it to be treated as a serious entry rather than a novelty.

What the cheap tier looks like from outside

From outside China, the interesting number is the 90 percent. It means the volume problem has been settled and the quality problem has not. A nine-in-ten AI share in a market that added hundreds of thousands of titles in eight months produces exactly the outcome you would predict: a flood of near-identical vertical dramas. That is why the recommendation lists, the eleven red lines and the "AI as tool" language all arrived in the same season.

Opera is one of the few places where the flood is not the main risk, because the material itself resists generic handling. A caicha melody is not a style prompt. The Susong dialect is not a tone descriptor. Either the model was trained on the real thing or the output sounds wrong to the only audience that would notice. The barrier is the training data and the performers who supplied it, not the generation step.

That is also why the project needed a broadcaster willing to fund something with no obvious viral hook. A seventeen-minute opera micro-drama is not an advertisement for anything. It exists because a provincial network with a heritage mandate decided the digital work was worth doing, and because the same network had already built the livestream audience to justify it.

An archive with distribution attached

Anqing Normal University is building a digital resource library, a virtual museum, a corpus for opera language and digital humans of its own. China Digital Culture Group used 5G and 4K to send a full "Niu Lang Zhi Nv" performance to public venues in several cities at once as a "second site" premiere, which is a way of putting a full production in front of people who will not travel to a theatre.

None of that is a business model yet. It is closer to an archive with distribution attached, and the funding comes from public cultural budgets rather than from ticket sales. That is a fragile arrangement in most countries, and it is the arrangement that produced the one AI micro-drama worth studying this autumn. The singing survives because somebody captured it properly, and for once the AI work was aimed at that rather than at volume.