As the first AI full-process theatrical film hits theaters, the income of grassroots actors and the number of productions starting in Hengdian have fallen sharply

Written by Juchao WAVE, author | Lao Yuer, editor | Yang Xuran

October 23, 2026, is destined to be a milestone worth remembering in China's film and television industry.

On this day, the first domestic full-process AI theatrical feature film to obtain a public screening license from the China Film Administration, "Sanxingdui: Future Past," officially landed in cinemas nationwide. In its 100 minutes of hyper-realistic imagery, all characters are original digital figures, with not a single live-action actor appearing on screen.

Also on this day, the General Office of the State Council published the "Opinions on Adapting to the Development of Artificial Intelligence Technology to Help Build an Employment-Friendly Development Model," which explicitly proposed to "effectively mitigate the employment substitution impact of AI applications" and to "steadily and prudently grasp the scope, intensity, and pace of AI application in labor-intensive industries, minimizing the impact on practitioners."

On one side, AI has formally obtained an "ID card" for the theatrical market, and virtual characters created by generative AI have charged all the way from the small screens of short videos and short dramas into the core battleground of traditional film and television; on the other side, top-level design has begun to sound the alarm over AI replacing jobs.

Above this line lies the complex performing art that relies on life experience, emotional layers, and improvisational expression; below this line lies standardized performance labor that can be broken down, described, and repeated.

And the vast majority of grassroots performing arts practitioners stand below this line, bearing the most direct impact of the technological wave.

Chill

The chill in the film and television industry is felt even by stars at the very top of the pyramid.

In July 2026, Golden Bell Award Best Actor Mark Chao admitted in an interview with "People" magazine that he had received only 10 script offers in the first half of the year, whereas at the peak of his career he received far more scripts in a single month.

The same anxiety has also struck Jia Nailiang. In June 2026, in a livestream, he faced netizens' question of "why do you keep doing livestream selling instead of acting" and bluntly said the four words "no dramas to act in." This formally trained actor who once won the Huading Award described his awkward situation: "not qualified enough to play elders, out of reach to play idols," with fewer and fewer opportunities for directors to come knocking.

An even more notable scene of stars collectively "job-hunting" took place at the Weibo Movie Night in June 2026. Stars such as Dong Zijian, Liu Haoran, and Cheng Xiao cared less about winning awards or walking the red carpet, and instead frequently made overtures to directors and producers on site, "looking forward to more work."

If top-tier actors still have the exposure power under the spotlight and can fight for opportunities for themselves, then the extras at the middle and lower levels of the industry have no chance to speak out at all.

The most tragic change has occurred in Hengdian. In the first quarter of this year, the number of short-drama shoots started in Hengdian plummeted by three-quarters year-on-year, and the occupancy rate of soundstages, once so scarce that a stage was hard to find, approached 80% vacancy.

At the peak in 2025, extras lined up at the actors' guild before dawn, with more notices than they could take; by 2026, the guild entrance is often empty. Official data show that Hengdian has cumulatively registered more than 150,000 "Hengpiao" migrant extras, but only 700 to 800 valid notices for ordinary extras are released each day.

China Newsweek reported in August that more than 60% of grassroots actors were in a near-zero-income state after the Spring Festival; some switched to delivering food, some returned to their hometowns, and others simply set up stalls on the streets of Hengdian selling snacks.

More noteworthy is that the short-drama industry, which was still a boom two years ago, has also taken a sharp turn for the worse.

Previously, top short-drama actors could command daily pay of 50,000 to 100,000 yuan, and mid-tier actors could steadily earn 1,500 to 2,000 yuan a day. But entering 2026, the daily pay of top actors directly fell to around 10,000 yuan; mid-tier actors' daily pay was cut to 500 to 800 yuan. The industry is contracting far faster than imagined. One leading short-drama company signed more than 200 actors in 2025, and by 2026 had terminated contracts with more than 100, with most of the remaining actors in a semi-unemployed state.

In March 2026, Yaoke Media, which had produced leading live-action dramas such as "An Jia" and "Jiu Chong Zi," suddenly officially announced the signing of two AI digital artists, Qin Lingyue and Lin Xiyan, and simultaneously launched the AI series "Qinling Bronze Mysteries." These two digital artists not only have their own film and television works but also opened social media accounts on Douyin, Xiaohongshu, and other platforms, updating their daily lives and taking on commercial collaborations like real artists.

The number of Chinese AI short dramas has shown explosive growth. Data released by the China Netcasting Services Association show that in the first quarter of 2026, about 128,000 micro-short dramas were launched across the industry, of which AI micro-short dramas reached 122,000, accounting for more than 95%.

Crushing

Previously, the production model for traditional live-action short dramas was already mature: the cost of a single title generally ranged from 500,000 to 1 million yuan, a crew typically had 40 to 50 people, and the entire cycle from casting, set building, and shooting to post-production took about two to three months.

Technological maturity has directly shattered the cost of short dramas. The lowest-end assembly-line AI short dramas in the industry can produce 80 episodes with a cost of 3,000 to 5,000 yuan, a team of 3 people, and 5 days; for premium-oriented AI short dramas, the cost is mostly between 100,000 and 200,000 yuan, the production cycle is 7 to 14 days, and the team needs only 3 to 8 people.

For "Zhanxiantai Live-Action AI Version," which went viral earlier this year, Xiong Binghui, CEO of the producer Kemeng Intelligence, gave clear figures: a 12-person team, a 30-day production cycle, and 100,000 yuan in computing power costs, with the overall production cost only one-quarter that of a live-action short drama of the same volume. If hidden costs such as coordinating actors' schedules, venue rental, and reshoots are included, AI's cost advantage is even greater.

Behind the dramatic drop in costs is that the technology side has crossed the key usability threshold.

From the end of 2025 to the beginning of 2026, mainstream video generation models made collective breakthroughs in character consistency, lip synchronization, and camera scheduling. In particular, in strongly fictional genres such as xuanhuan, apocalyptic, and period drama, they can already deliver stable output, and the visual quality has reached a level that is "usable for short dramas."

When technology is no longer an obstacle, capital and platforms will naturally vote with their feet and shift to production methods with higher efficiency and lower costs.

It can even be said that in the vast majority of short dramas, the functional roles of characters are almost fixed: the villain is responsible for being arrogant and getting slapped down, the subordinates are responsible for nodding in agreement, the elders are responsible for urging marriage and applying pressure, and passersby are responsible for gasping in amazement and egging things on. They do not need complex inner drama, nor do they need richly layered emotions; they only need to stand in the designated position, finish the designated lines, and complete a twist or a punchline.

What these short dramas output is in fact the alienation of acting into a mechanical, repetitive form of performance labor. In AI's logic, such labor can be clearly broken down into three modules: action, expression, and dialogue, and can be described in a standardized way and repeated an unlimited number of times. "Can be described in a standardized way" is precisely what generative AI is best at.

Under such circumstances, replacement is almost inevitable.

Space

But it is very difficult for it to replicate the layers of complex emotions, and even more impossible for it to capture the various subtle manifestations in improvised reactions. Just as it cannot accurately perceive the inner turmoil beneath Shurou's calm expression when she learns that her husband has remarried, nor can it replicate Xu Fu's emotional outburst after witnessing the death of a close friend. These are the true barriers of acting as a profession.

Take the translation industry, for example. Before the spread of generative AI, the translation of ordinary business correspondence, product manuals, and everyday informational texts was the main source of income for a vast number of junior translators. Today, AI's translation accuracy for standardized texts has exceeded 95%, the cost per thousand characters is less than one-tenth that of human translation, and delivery efficiency is dozens of times higher.

The "2026 China Translation Industry Development Report" released by the Translators Association of China shows that AI's replacement effect on junior translators is obvious. But at the same time, the report also points out that enterprises still account for more than 40% of demand for intermediate- and senior-level translation talent, and "senior-level talent, because of its irreplaceability, has instead become scarcer."

Take the customer service industry as another example. Gartner survey data shows that by 2026, conversational AI will already be able to help call centers reduce labor costs by about 80 billion US dollars. For standardized entry-level customer service positions such as order inquiries, password resets, and business consultations, the replacement rate has already reached 85% to 90%.