In China, AI-generated video is not merely threatening actors, influencers and livestreamers; in some productions it has already replaced human performers. The turning point was the release of ByteDance's Seedance 2.0: digital performers can now produce better video faster and more cheaply than people.

The numbers show the pace of the shift. About 128,000 short dramas were published in China in the first quarter — three times the total for all of the previous year on its own. According to the China Netcasting Services Association, 95 percent of them were AI-generated.

The cost gap

The figure Tsinghua University professor Shen Yang gave the Financial Times: a minute of AI video now costs only $90 to $120. That is roughly ten percent of what production with human actors used to cost.

A gap that size stops being a matter of preference within an industry. The method that is ten times cheaper becomes the only method the moment quality is equal — and with Seedance 2.5 and Wan 3.0, video quality has jumped again.

The size of the affected group

  • 690,000 people are directly employed in the industry.
  • 15 million list livestreaming as their primary job.
  • AI-related labour disputes have risen over the past two to three years, according to lawyers.

A caution about the figure

The 95 percent number is the industry association's own measurement, and where "AI-generated" begins is not specified. A production made entirely by a model and one where only backgrounds or voices are generated may fall into the same bucket. The direction is not in dispute, though: total output tripled over the same period, and that volume is simply too large to be produced by human crews.

The most unsettling detail

One sentence in the reporting captures the whole argument: some performers are being forced to "distil" their voice and likeness into AI tools before being fired.

The person does not merely lose the job; they train the system that takes it. The reason they are not needed on the next production is the data they handed over on the last one.

This differs from the familiar shape of automation debates. A machine on a factory floor replaced a worker without taking anything from the worker. Here there is a handover: the thing that replaces you learns from you.

Why here first

The short drama format is ideal ground for this shift. Episodes are measured in minutes, production volume is enormous, viewer expectations are not cinematic, and the content is consumed and forgotten quickly. It is the place where the quality bar is lowest and the volume is highest.

So the picture in China is not an exception but a preview. AI video takes the cheapest, most mass-produced format first, then works upward as quality rises. For anyone wondering when a given job is affected, the measure is simple: the lower the cost per unit, the earlier the turn comes.