Chinese Companies Clone Actors' Faces with AI, Fire the Humans—95% of Short Dramas ...

TEXT START: In China, a growing number of companies are training artificial intelligence on actors' and online influencers' faces, voices, and performance styles to create digital replicas—then firing the original humans.

The Dissection

The text documents a clean instance of labor being converted into reusable AI capital. The worker supplies the training data, then becomes redundant once the replica can perform indefinitely at a fraction of the cost.

Its strongest evidence is not the dramatic 95% statistic. It is the mechanism: train once, reproduce continuously, sever the human. That directly supports P1 and shows P3 emerging in content production. The article also exposes the next battleground—ownership of faces, voices, expertise, and performance data.

The promotional closing about BigGo Finance appearing in Google Search Top Stories is SEO residue, not analysis. It weakens the article’s editorial credibility.

The Core Fallacy

The text treats the transformation primarily as a labor-rights and compensation dispute. That is too small a frame. Severance, veto rights, and five-year licenses may alter who extracts value from a digital replica; they do not restore the worker’s productive role.

The decisive change is not that actors are being paid less. It is that their labor is no longer required after their identity and performance have been distilled into capital. Contractual protection can redistribute the carcass. It cannot revive the employment-to-wage-to-consumption circuit.

A secondary evidentiary flaw is the unexamined equation of “AI-produced” with fully AI-generated. The text provides no methodology for the 95% figure, cost comparisons, or quality controls. The exact percentages may be inflated or loosely defined; the displacement mechanism remains structurally clear.

Hidden Assumptions

• Demand for content will expand enough to absorb displaced workers.
• Workers can simply move into “other lines of work” without encountering the same automation.
• Legal rights over likeness and voice will remain enforceable and economically valuable.
• Temporary licensing, veto rights, and severance can preserve bargaining power after replacement.
• AI output is sufficiently acceptable to consumers across the cited categories.
• Reported production costs include all meaningful costs, including infrastructure, supervision, distribution, and correction.
• The 95% and 10% figures use stable, comparable definitions.
• Companies will share productivity gains rather than concentrate them among owners of models, platforms, and distribution.
• The affected workers can become owners or controllers of the AI systems replacing them.

The most important smuggled assumption is that displaced people remain economically necessary because the industry still exists. That is false under the Discontinuity Thesis. An industry can grow explosively while its labor force collapses.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The article accurately displays the replacement process and the emerging legal conflicts. But by centering severance packages, usage periods, and veto rights, it makes systemic dispossession look administratively manageable. The audience is shown the burning building and handed a better lease agreement.

Its language of resignation also performs normalization. “Unavoidable” converts a competitive economic result into weather. The firms’ ownership of cloned identities and accumulated expertise disappears into the background, where the real power sits.

The Verdict

This is an early, highly visible autopsy of the post-WWII labor circuit. Chinese short dramas and livestream commerce are demonstrating the sequence in miniature: cognitive and performative labor is digitized, production costs collapse, output scales, and human participation loses necessity.

The article does not prove that all cognitive labor has reached terminal replacement, nor does it establish P2 across the whole economy. It does show P1 and P3 operating in a live market. Legal friction, cultural resistance, quality limits, and rights disputes are lag defenses—hospice care, not reversal.

The humans are not merely losing jobs. Their faces, voices, and expertise are being retained as corporate assets after their bodies are dismissed. That is the structural fact the article approaches but does not fully state: the worker is being replaced by a copy of their own accumulated labor, while ownership of the copy determines who survives.