AI Short Drama Prices Collapse in China—but Finished Shows Still Cost More

A September 6 report from 21st Century Business Herald, citing CCTV Finance, says quotes for standardized AI short-drama production in China have fallen from about RMB 5,000 per minute to several hundred yuan. The same report puts customized work at RMB 10,000 to RMB 20,000 per minute. That gap is the real signal: the market is separating repeatable output from projects that still require original direction, controlled characters, revisions, and client-specific delivery.

These figures are vendor quotes reported in China, not a global price index or a rate card used by every studio. “Several hundred yuan” is also a range without a disclosed median, sample size, resolution, episode length, acceptance standard, or list of included services. It should not be presented as the average cost of a finished AI microdrama.

• Read the September 6 cost report ↗

The lowest quote can describe a standardized pipeline while excluding the work that turns generated clips into an accepted episode. A production still has to absorb rejected generations, continuity repair, editing, audio, captions, rights review, project management, and delivery. Marketing and distribution sit outside production again.

A March investigation by Cailian Press illustrates the scope problem. Its interviewees estimated compute at RMB 1 to RMB 3 per generated second before two to five revision rounds, while complete photoreal AI-drama work was described at roughly RMB 2,000 to RMB 6,000 per finished minute and premium work above RMB 10,000. The same report gave a typical delivery window of 35 to 40 days. Those numbers do not disprove the newer low quote; they describe a more demanding product and show why every cost claim needs a defined unit.

• Model retries and cost per usable finished second →
• Read the Cailian Press production-cost investigation ↗

The September report cites a Zhongtai Securities estimate that a conventional short-drama project often uses about 15 people and takes more than four weeks, compared with three to five people and roughly 10 days at the current AI-industry average. It also says production costs across 2D, 3D, and photoreal AI categories fell 80% to 90% from 2025 levels and reached 2% to 10% of traditional short-drama costs. These are analyst estimates spanning different formats, not an audited before-and-after study of identical shows.

A separate CCTV report from Zhengzhou provides a useful case rather than a universal benchmark: one company said a 10-person team could complete a 120-minute photoreal animated drama in seven to 10 days, with per-minute cost falling into the hundreds of yuan. Together, the reports support a direction of travel—smaller crews and faster iteration—but not one standard crew size, schedule, or quality level for every AI production.

• Follow ReelPulse's controlled production workflow →
• Read CCTV's Zhengzhou production case ↗

DataEye lists a July 24 report covering China's AI drama and comic-drama sector. Figures repeated from that report put the market above RMB 22 billion in the first five months of 2026 and forecast about RMB 40 billion for the full year, up 138%. The category includes AI comic dramas and other generated formats; it is broader than photoreal AI short drama and should not be compared directly with app-store revenue or a production-services market.

The prediction of one million Chinese AI short dramas in 2026 is attributed to Zhongtai Securities. It is a supply forecast, not a completed-title count. The same caution applies to audience figures: viewers, platform accounts, payers, and cumulative reach are different measures. For overseas figures, RMB-language reports describing USD 4 billion should not be mistranslated as $40 billion. Definitions and units matter more as the numbers grow.

• See DataEye's official report index ↗
• Compare the defined app-store market scope →

There is no public evidence that ReelShort, DramaBox, DramaWave, or another global app buys finished shows at the quoted Chinese standardized rates. They are distributors, commissioners, and marketers with different production arrangements. Attaching the RMB figures to a named platform without a contract, rate card, or company disclosure would turn an industry signal into an unsupported company claim.

Cheaper output expands the number of concepts that producers can test; it does not make audience attention cheaper. One producer quoted by CCTV in August said that during the 2026 Spring Festival window, live-action titles were only one-fiftieth as numerous as AI titles but generated 25 times the total plays. That is one reported comparison, not a category-wide measurement, but it captures the commercial problem: abundant supply can coexist with weak demand.

The durable advantage moves downstream to scripts, character continuity, recognizable intellectual property, localization, distribution, and retention. ReelPulse's working rule is therefore simple: price the submitted output, log every failed and rejected attempt, calculate the usable finished seconds, and publish the evidence. Generation is becoming a commodity; an accepted story is not.

• See the Community Cut evidence protocol →
• Read CCTV's live-action and AI audience comparison ↗