Lost in translation? AI helps Chinese micro-dramas go global

In Chinese imperial dramas, high-ranking women often refer to themselves as bengong, literally "this palace". The translation is likely to baffle audiences unfamiliar with China's ancient royal courts, so translators typically use "I" instead and let titles, tone, and context convey the character's royal status.

Solving such problems has become a growing business for Transn IOL Technology, an artificial intelligence company based in Wuhan, Hubei province. The company has translated and dubbed more than 20,000 Chinese micro-dramas for markets including North America, Southeast Asia, and the Middle East.

Micro-dramas, typically told in bite-sized episodes with cliffhangers, are drawing a growing international audience. The micro-drama market outside China is expected to exceed $6 billion this year, according to DataEye Research Institute. Global downloads of short-drama apps surpassed 850 million in the first quarter of 2026, up 140 percent year-on-year, according to market intelligence firm Sensor Tower.

Transn previously provided translation and dubbing for China Central Television productions. It has since built a workflow that combines automated translation, human review, and voice cloning. Its TaiHow Video platform separates dialogue from background sound, translates scripts in context, generates multilingual dubbing, and reassembles the audio.

The process runs on the company's proprietary large language model, which is designed for the high volume and fragmented format of micro-dramas, said Fu Qiang, vice-president of Transn.

"AI first handles the bulk translation, then multiple AI agents cross-check the results—essentially machines checking machines—before language experts conduct two rounds of human review," Fu said.

According to Shi Xin, vice-president of Transn, the system can process an hour of video content in approximately the same amount of time at the automated stage. With human review, a version for market testing can be delivered within one or two days, and a polished production in three to five. A fully manual process takes at least two weeks.

The company says AI has cut its overall translation and dubbing costs by around 30 percent, but barriers to success remain. Revenue from its film and cultural translation business grew 29.8 percent year-on-year in the first half of 2026, according to its interim report, but its gross profit margin fell by 13.19 percentage points.

Human translators still handle what machines cannot. Chinese programs frequently feature cultural concepts that have no straightforward equivalents in other languages. A "live-in son-in-law", for example, carries expectations about family relationships, and the elaborate ranks of imperial consorts require consistent terminology throughout a series. Reviewers correct cultural misunderstandings and make sure characters' personalities and emotions survive the translation process.

Tastes also differ by market. North American viewers favor romance and fantasy, family-centered stories resonate in Southeast Asia, and Middle Eastern audiences show interest in romance and inheritance-related plots, Shi said.

For Transn, the next challenge is making translated stories feel natural to local audiences while maintaining the speed needed in a fast-moving entertainment market. AI can help a Chinese drama reach a foreign viewer's screen much faster. Whether its next plot twist keeps that viewer watching still depends on how well the story is told.