An AI short drama titled Menghui Honglou (《梦回红楼》, Dream Back to the Red Chamber) rebuilt a national television classic with generative-video tools. It used material from the 1987 series Dream of the Red Chamber as its base, composited the facial features of 17 of that series' actors onto AI characters, rewrote the plot into confrontation scenes the original never contained, and ran on a Douyin account signed to MCN Qichuang Media, a talent-management firm. Its episodes carried no "AI-generated" label at any point. The hashtag hit the hot-search list on Nov 12, and the drama came down that same evening. [1] [2]
For anyone producing video with generative tools, this case is where four separate legal fronts converge on a single production. A lawyer quoted in one of the reports counts four: the actors' portrait rights, the personality interests of a deceased actor, performers' rights, and the original producer's rights of reproduction, adaptation and integrity. [1] The portrait and deceased-person fronts carry a statute in force since 2021 and a court-tested identifiability standard; the labeling front carries dated rules that took effect before this drama shipped; the performers' and producer's fronts rest, on the public record so far, on that lawyer's stated claims. The working answer is to inventory every borrowed element — faces, material, performance, plot — before publishing, because each element now carries its own claim, its own rulebook, and a visible consequence.
An identifiable face is actionable, even one from 2007
Article 1019 of the Civil Code bars any organization or individual from infringing portrait rights by defacing, defiling, or falsifying a person's image through information technology, and bars making, using, or publishing a person's portrait without consent. [3] The question courts actually ask is identifiability. In a case note published by the Shenzhen Nanshan district court, a vlogger's video was used as a face-swap template by an app operator; the court held that even in period costume and hanfu styling, an ordinary viewer could identify her from her facial and physical features, and that synthesizing a swapped face destroyed the identity link between image and subject, which constituted infringement of her portrait right. [4]
Lawyers applied the same standard to this drama: once an AI-composited image is identifiable as a specific natural person and is used without permission for commercial traffic, it constitutes portrait infringement, and an "AI fictional character" is no shield. [2] The 17 names include Zhang Li (Baochai), Deng Jie (Wang Xifeng), Gao Hongliang (Jia Lian), Li Ting (Grandmother Jia), and Chen Xiaoxu, the Lin Daiyu of the 1987 series, who died in 2007. [2] Death does not end the claim. Article 994 of the Civil Code lets a deceased person's spouse, children and parents demand civil liability for infringements of the deceased's name, portrait, reputation, honor, privacy or remains, with other close relatives as fallback claimants. [3] Per the reports, Zhang Li, Ouyang Fenqiang and Deng Jie have formally filed suits seeking removal of the content across platforms. [2]
The statute's exceptions for using a portrait without consent cover personal study and research, unavoidable news reporting, official duties, displaying specific public environments, and public-interest acts. [3] None of those categories describes a commercial drama series built for traffic.
The borrowed material pulls in performers and the original producer
The faces are one front. The lawyer assessment also names performers' rights, plus the original production company's rights of reproduction, adaptation and integrity [1] — the kinds of claims that presuppose use of the original work, and the drama is built on the 1987 series' material with a rewritten plot. The portrait article adds a detail that matters here: even the rights holder of a portrait work, the party holding the 1987 material, may not publish, reproduce, distribute, rent or exhibit the portrayed person's image without consent. [3]
The account behind the drama had published 368 Red-Chamber-related works, 21 of them AI episodes; the top-liked AI episode reached 115,000 likes, against the low hundreds its clip compilations used to draw. [2] The rewrites were announced openly in episode titles; the report describes an episode-16 title reading the princess's fury with Baochai slapped in the street, and a plot in which Lin Daiyu is made a princess whom Grandmother Jia kneels to while Baochai and Tanchun are humiliated in turn. [2] The performers-and-producer claims are lawyer-stated positions from the reports, a different evidentiary grade from the statutory text and the court-tested portrait claims above.
The labeling rules took effect before this drama shipped
Face-swapped drama is deep synthesis under Chinese rules that predate this case. The Deep Synthesis Provisions, issued by the cyberspace and public-security regulators jointly with the industry ministry and effective Jan 10, 2023, require prominent labels on face replacement, face generation and person-video editing services where public confusion is possible, and prohibit any organization or individual from technically deleting, tampering with or hiding those labels. [5] The same provisions require providers and technical supporters offering face or voice editing features to prompt users to inform the person being edited and obtain their separate consent [5] — an obligation the rules place on the tool providers; the consent duty binding a drama producer comes from the portrait article itself. [3] The provisions define a "deep synthesis service user" as anyone who uses such services to make or publish information, a definition the drama's producers fit. [5]
The regime then tightened. The Measures for Labeling AI-Generated Synthetic Content, jointly issued by the cyberspace, industry, public-security and broadcast regulators on March 7, 2025 and effective Sept 1, 2025, require a prominent notice on the start frame of an AI-generated video and around the player, keep the label inside files that are downloaded, copied or exported, and obligate users to declare AI content and use the platform's labeling function when publishing. [6] Platforms must verify file metadata for hidden AI markers and add prominent notices for content that is confirmed, declared, or suspected to be AI-generated; maliciously deleting, tampering with, forging or hiding the labels is prohibited, as is providing tools for doing so. [6] For micro-dramas specifically, Cailianshe reports that the Measures for the Development of Micro-Dramas (《微短剧发展管理办法》), effective Sept 1, 2026, require AI-generated or AI-produced episodes to carry a notice label in a prominent position of every episode. [7]
The enforcement calendar brackets the case. The National Radio and Television Administration's current special governance runs from Nov 1 to the end of December, its third micro-drama action of 2026 and the first to bring AI micro-dramas and algorithm distribution systematically into a campaign framework, with "face theft" and "voice theft" named as governance priorities. [7] In January the regulator targeted videos that "AI-modify" the Four Great Classical Novels; in June it ran an eight-category cleanup that included piracy. [7] The Civil Code was adopted May 28, 2020 and has been in force since Jan 1, 2021. [8] [9] The drama's AI episodes went up starting Oct 18, the campaign began Nov 1, and the controversy broke Nov 12 — all of these rules were already in force when the first episode shipped. [2] [7]
Takedowns, suspensions and an 810,000-yuan claim are already on the record
None of the consequences attached to this production model is hypothetical. The drama itself was pulled the evening of Nov 12. [1] In March, the AI short drama Taohua Zan (《桃花簪》) was reported to have copied a hanfu blogger's styling and used her facial image for a villain role without authorization; Hongguo removed the drama entirely and suspended the producer's uploads for 15 days. [7] In the Shenzhen face-swap case, the court ordered the app operator to apologize and pay 5,000 yuan covering property loss and reasonable enforcement costs, a figure that became 5,250 yuan in the mediated settlement on appeal; the damages rule the court cited reads the victim's loss or the infringer's gain, with the court setting the amount where neither can be determined. [4]
The star cases run in the same direction. Yang Zi's studio declared it was preserving evidence the same day her image surfaced in a March drama; Yi Yangqianxi's studio moved within a day in April and the episodes later came down; Dilireba sued a face-swap company, and the court ordered a public apology and damages. [2] On Nov 10, Zhao Liying's AI face-merge suit against the Hongguo short-drama platform was heard in a Beijing court; per the report, she listed the platform operator and Douyin's parent company among the defendants and claims 810,000 yuan — a demand, not a court award. [2] A second report's headline corroborates the 810,000-yuan figure and adds further actors pursuing AI-related claims. [10]
What this changes before the next AI-video project
Each front carries its own grade of evidence: verbatim statute for portraits and deceased-person interests, a published court note for the identifiability test, lawyer-stated claims for performers and the producer's rights, and dated, checkable rules for labels. The consequences visible so far run from a 5,250-yuan mediated payment and a 15-day upload suspension to an 810,000-yuan demand; the suits in this case sit at the filing stage in the reporting, with nothing yet adjudicated. [2] [4]
For the next generative-video project, the record supports a short set of working assumptions. Treat an AI-composited face as actionable once an ordinary viewer can identify the person behind it, whatever styling changes were applied. Treat the interests of deceased performers as claimable by their families. Expect the start-frame label, the metadata markers and the user declaration duty to apply at publish time, and label-stripping to be separately prohibited. Expect platforms to check metadata and to carry their own verification duties. It is not legal advice. The labeling rules do not substitute for the portrait article's consent requirement; a composited face still needs the identifiable person's permission before the video ships.