In the Age of AI, Is Chinese Opera Being Revived or Reduced?
Gong Yan This generation’s first encounter with Chinese opera could hardly be more different from those of the past. Instead of spacious theaters or TV channels dedicated to opera, young people are far more likely to swipe past an “AI Peking opera costume change” video on Douyin, watch an “AI opera epic” on Bilibili, buy a Yue opera skin in a video game, or listen to a traditional opera tune sung by an AI singer.
Chinese opera is expanding beyond its traditional theater spaces and into a new media environment, a transformation that goes far deeper than a simple change of medium. In the past, Chinese opera relied on the trifecta of troupes, famous actors, and live performances, with its heart nestled in physicality, time, and the theater-going experience. On today’s social media platforms, AI video generation tools, and digital industries, however, Chinese opera is being broken down. Its voices, postures, face masks, and movements have become datasets and visual references for algorithms, distribution platforms, and generative AI systems.
Since 2024, this digital digestion has only accelerated with the proliferation of generative tools like Midjourney, Runway, Kling AI, Dreamina AI, and Sora. The average user needs only to enter simple prompts such as “Peking opera ingénues,” “Kunqu opera actress,” “cyberpunk Chinese opera,” or “Chinese opera stage,” and AI will automatically generate characters, scenes, and even videos styled after the great operatic variants. What once required professional animation teams, industry infrastructure, and costly production processes can now be replicated by ordinary creators with remarkably low barriers to entry.
One key precondition for this transformation is Chinese opera’s highly conventionalized nature. Unlike modernist theater, traditional Chinese opera has long been built upon highly structured foundations. From the traditional role system of Chinese opera consisting of four major role types: sheng (male roles), dan (female roles), jing (painted-face roles), and chou (comic roles), to performance structures based around singing, chanting, acting, and combat, and even the templates governing face masks, costumes, and water sleeves (long flowing sleeve extensions used in Chinese opera performance to convey emotion and gesture), nearly every aspect of Chinese opera is highly regimented and symbolic. Choreographed actions like “posing,” “stage circling,” and “donning armor” in Peking opera are repetitive and visually distinct, and the internal logic of face mask colors — red for loyalty, white for cunning, black for integrity — makes them reliable visual markers.
The conventionalization of Chinese opera, originally shaped through centuries of aesthetic experimentation, has become generative AI’s gateway to decoding the art form. At its core, generative AI relies on quantitative training, stylistic identification, pattern recognition, and image association. Art forms that are visually distinctive, regimented in their movements, and stylistically identifiable are more easily digestible to its algorithms. Chinese opera checks all of those boxes.
For the first time in its history, Chinese opera is being systematically broken down into trainable, adjustable, and generative data models through AI. Water sleeve movements, once embodied solely by Kunqu opera performers, are now being transformed into digital trajectories through motion-capture systems. Face mask designs, once a hand-painted tradition, have now become visual tags for image generation. And Chinese opera singing voices, transmitted orally from master to student, are now being deconstructed into frequencies and audio compositions by AI voice systems. Most importantly, Chinese opera has become accessible to creators on an unprecedented scale. Though elements of Chinese opera have already entered the public consciousness through film, television, and video games, its production has historically been relegated to corporations and professionals. For example, it took teams of artists, motion capture experts, 3D modelers, and extensive post-production work to create Chinese-opera-inspired characters in games Arena of Valor and Sword of Justice. There was no room for the average creator to participate. AI has taken the gate separating that average creator from Chinese opera and thrown it wide open.
Since 2025, AI-generated opera manhua (Chinese comic-style videos) have proliferated across the internet. The Pan Long Bangzi theater troupe based in Jinan, capital of the eastern Shandong province, for example, released “Judge Bao on the Case,” a reworking of traditional Bangzi opera native to northern and eastern China that combines AI, Chinese opera, and manhua, reaching 5 million views on Douyin, the Chinese version of TikTok. The series comprises 100 episodes, each 30 seconds long. Only some contain traditional Bangzi singing; the rest are mostly AI-generated animated sequences. While some viewers felt this format diluted the opera’s essence, its reach is undeniably broader than that of traditional opera recordings. The AI shorts on Daxi Henan — a Douyin account about all things Yu opera, a form of Bangzi opera specific to the central Henan province — featuring 3D-animated characters singing Yu opera tunes inspired by famous television segments, has reached more than 18.63 million views. AI Chinese opera content like this moves beyond straightforward opera viewing and evolves into secondary creations based on classics while still achieving substantial reach.
“Chang’an Kitchen God: Year of the Horse Collection,” published through a collaboration between the northwestern Shaanxi province-hailing Xi’an Sanyi Qin Theater Troupe and Xiying Media, combines AI-generated scenes, Qin-style tunes, the Shaanxi dialect, and vertical short videos to create an entirely new format. In the series, the Kitchen God and his spouse ride through both ancient and modern Chang’an (now known as Xi’an, the Shaanxi capital) on an AI-generated “flaming horse.” In addition to short videos, artificial intelligence is increasingly being used to create opera-themed short films, interactive games, and other forms of digital content, further expanding the ways in which traditional opera is created, experienced, and shared.
AI is redistributing power to produce Chinese opera content, shifting it away from official troupes and professional film studios toward local theaters, student teams, and individual creators. But the impact of AI on the dissemination of opera goes beyond professional content production; it has also entered the everyday media practices of ordinary users. On platforms like Douyin and Xiaohongshu, also known as RedNote, interactive features such as “AI opera dress-up,” “AI opera face-swapping,” and “upload a photo to generate an opera character” have become widespread. Users simply upload a selfie or activate their phone camera, and the system automatically generates images of Peking opera ingénues, Yue opera lead female roles, or Chinese-opera-style characters. Some AI mini-programs can even transfer opera makeup, adapt facial expressions to match a role, and generate opera costumes, quickly “immersing” the average user in the visual world of Chinese opera.
Yet at the same time, a deeper problem is emerging. While AI can learn the aesthetics of “Chinese style,” “Eastern beauty,” or “operatic visuals,” it does not truly understand the literary structure of “The Peony Pavilion,” nor can it grasp the bodily rhythm and timing of a Kunqu performance. As a result, Chinese opera is increasingly being compressed into a visual style, a platform hashtag, a repository of culturally coded Chinese audiovisual material, or a dataset in an AI training library.
The reality is that platform algorithms are most eager to spread opera’s most visually striking elements: swirling water sleeves, the grand entrance of an ingénue, the phoenix crown and embroidered robe, the highly saturated “Chinese red,” and the climactic high notes of operatic singing. These fragments are far more likely to receive clicks and recommendations than complete operatic texts. As AI continuously reinforces these visual markers, it further shifts opera from a “text” towards an “aesthetic.” What AI is truly good at is not the operatic text itself, but the efficient replication of the “opera aesthetic.” In the age of AI, what is being widely disseminated may no longer be traditional Chinese opera itself, but rather symbols, “distilled” algorithms, and technology.
Translator: Gabriel Kwan.
*(Header image: Visuals from N**ovelquick, *WeChat, and VCG, reedited by Sixth Tone)