Chinese microdramas have become a testing ground for generative video faster than feature films. Short vertical episodes are easier to produce in series, and viewers watch them on small screens. In the first quarter of 2026, around 128 thousand new microdramas were released in China, of which approximately 122 thousand were fully generated by AI.
A share above 95% looks like an almost complete replacement of filming. But this figure counts released titles, not views, revenue, or quality. To understand the real shift, the figure needs to be broken down into production, economics, and format limitations.
What exactly was counted in 95%
The data refers to the first quarter of 2026 and comes from a report by the China Netcasting Services Association. According to the Channel News Asia analysis, around 122 thousand of the 128 thousand releases were created without cameras or live actors. Projects filmed by people using AI for subtitles or postproduction were not included in this group.
So this really refers to fully generative works. But the unit remains a release. One cheap series and one major project carry equal weight in the statistics. The metric does not answer three important questions:
- how many productions viewers watch to the end;
- what share recoups its costs;
- how much time viewers spend with AI and live-action content.
A large number of releases primarily indicates a low production threshold and a high speed of story testing.
Why the microdrama suits generative video
Microdramas consist of short episodes, often one to two minutes long, and are designed for a vertical screen. The camera usually stays close to the character, the editing is fast, and the plot is built around frequent twists and cliffhangers. This format conceals some of the weaknesses of generation.
Models can already create a convincing individual frame or short shot. It is harder to preserve spatial logic, lighting, clothing, and character movement throughout a long scene. A narrow frame reduces the amount of background, while frequent editing makes it possible to replace an unsuccessful fragment without reworking an entire episode.
Production is also easy to break down. The script is converted into scenes; references are prepared for each one; several takes are generated; the successful ones are selected and assembled in the edit. AI speeds up individual operations, but selecting takes and maintaining a consistent style remain manual work.
How many people and how much time are required
Practical estimates vary considerably. In a CNA report, a producer at a Chinese studio says that a team of five to ten people and between two weeks and a month are needed for a complete series. One project was estimated at approximately 100 thousand yuan, whereas a live-action costume microdrama could cost 300–500 thousand yuan and take two to three months.
This is the experience of specific market participants, not a universal rate. The price varies depending on length, number of characters, voice-over quality, number of repeat generations, and platform requirements. The closer the result is to professional video, the more failed takes have to be regenerated and edited.
Not all of the work becomes cheaper. The filming location, transportation, and some extras disappear, but expenses arise for computing, version storage, continuity control, and rights to source materials.
Where savings turn into rework
A beautiful individual shot sells the technology better than a finished series. Viewers notice other flaws: a character changes between shots, an object disappears, the direction of movement shifts, or the voice does not match the emotion. In a long story, errors accumulate.
A representative of iQiyi noted that professional long-form storytelling remains an industry-wide problem. In one scene, the weather, lighting, or movement of the waves may look different from one shot to another. Correcting such discrepancies requires new generations and increases computing costs.
That is why it is incorrect to calculate production cost based only on the price of the first generation. A full production budget is needed:
- script and storyboard development;
- character preparation and visual rules;
- a package of repeat generations;
- voice-over, music, and mixing;
- editing and continuity checks;
- legal review of likenesses and source materials;
- adaptation to platform requirements.
How professions are changing
Demand for repetitive production operations is shrinking fastest. In the report, market participants mention extras, basic lighting, low-cost filming crews, and purely technical editing. A team that previously employed five or six editors may be reduced to one specialist because generation already produces material with some of the cuts assembled.
At the same time, a new role is emerging: a person who repeatedly generates variations and selects usable shots. This is neither quite a director nor a traditional editor. The work is closer to managing probabilistic production, where an exact take cannot be ordered in advance.
Creative decisions do not disappear. Someone still has to maintain the character’s personality, the story’s pace, and its visual language. The more raw material the system produces, the more important selection becomes. A number of options without an editorial decision quickly turns into a monotonous stream.
Rights to a face and a digital twin
Generative production is creating a separate market for licenses to a person’s appearance, facial expressions, and voice. A studio may reach an agreement with an actor to use their digital likeness in new scenes without their physical presence. This is not a single permission “for the face,” but a set of conditions.
The contract should define:
- specific projects and genres;
- the term and territories of use;
- the right to create new lines and actions;
- the possibility of transferring the model to contractors;
- the procedure for approving disputed scenes;
- storage of original scans and training data;
- deletion of the digital model after the license expires.
Without these boundaries, consent for one project can turn into uncontrolled reuse. Viewers also need clear labeling; otherwise, it is impossible to distinguish an actor’s real participation from a licensed digital copy.
Why there are many releases but few hits
Low costs increase the number of attempts. A studio can quickly test dozens of stories and keep those that receive a response. But viewer demand does not grow at the same speed. Platforms receive thousands of similar stories competing for the same attention.
A Chinese industry publication, citing association data, describes a shift from increasing quantity to searching for higher-quality projects. This is an expected stage: technology first sharply reduces the cost of release, then the bottlenecks become the script, a distinctive style, and distribution.
The 95% figure itself is easy to misread. It does not prove that audiences prefer AI content or that live actors are no longer needed. It shows how quickly supply responds to an inexpensive tool.
What can be applied to other content projects
| Lesson from microdramas | Practical takeaway |
|---|---|
| The short format conceals some defects | Start with tasks where errors are localized and fragments can be easily replaced |
| Release volume grows faster than demand | Measure retention and recoupment, not just the number of publications |
| Generation makes the first version cheaper | Calculate the cost of selection, rework, and rights checks |
| Repetitive roles are being reduced | Move specialists into quality control and editing |
| A digital likeness becomes an asset | Define the scope of the license and how the model will be deleted |
Chinese microdramas show not the end of the film industry, but the limits of factory-style generation. Creating thousands of releases has become easier. Making a story that viewers will watch to the end and remember remains difficult.
Compare models before you start
The service sets its plans, limits and model catalog. If they differ from this article, contact us so we can update it and record a new review date.
Browse modelsAffiliate link: your price stays the same and the project earns a commission.
