A long 36Kr feature, first reported by the Chinese tech outlet All-weather Tech, argues that AI-generated film and television in China is leaving the lab and entering the formal industry. Its evidence comes from festival rules, prime-time television and the feature-film plans of the big streaming platforms. The creators it interviews expect a wave of long-form AI releases to reach audiences within months, and the platforms are already setting up the money and review systems to carry them.
The clearest sign is at the festivals. On 14 September the China Golden Rooster and Hundred Flowers Film Festival opened entries for its first standalone AI film brand. The rules differ from any ordinary submission: at least half of every entry's footage must be AI-generated, features must run 60 minutes or more, and entrants must hand over their model toolchain, core prompts, a description of the generation process and engineering screenshots so the jury can check how the work was made.
Other events got there earlier. The Beijing and Shanghai international film festivals already run AIGC sections, and the Hundred Flowers Awards added an AIGC showcase this year. Broadcast has moved too. At the end of August, Hunan TV and Mango TV put an AI series, Post Journey to the West, into a prime-time slot, airing five episodes of a 30-episode run with the rest due over the winter holidays. The article calls it the first time long-form AI content has reached that kind of traditional TV slot.
To show why this matters, the piece starts from the economics of a traditional set. A shoot is collective labour tied to people, money and space: cast and departments must turn up together, sets have to be built for houses that do not exist, and a few minutes of rain can need water trucks, lights and dozens of crew. Every day the machine runs costs money, and every unfinished scene pushes the schedule back.
The biggest change the feature describes is to the crew. Commercial director Liu Chi, whose six-person team made the 45-minute animated film Floating Life in Seven Days, says she has seen nothing so disruptive in more than ten years in the business. That film won best feature in the AIGC section of this year's Beijing festival after about six weeks of work, roughly three of them on generation and post. For the theatrical project she is on now, more than ten people work on AI generation alone, yet the whole team, writers and post included, is 20 to 30 people, against the hundred or so a conventional shoot might need. There is no location shooting at all; she likens the operation to a post-production house.
Director Chen Youxue describes the new process as something between live action, animation and post-production. Directors use text commands to drive 3D models in Blender and lay out blocking, camera positions and moves. Actors still perform, and their movement and expressions are carried over onto the AI characters. Work that once happened on sets and under lights becomes digital asset building and shot planning. In principle, once the character assets are settled, adding people and computing power lets a team generate material in parallel and squeeze the schedule.
That does not mean films come out of a button press, and none of the interviewees claims the machine can judge a story for them. Liu still works through her film scene by scene, checking that characters do not drift and that the emotion holds together. Long Yitong, a Central Academy of Fine Arts student who began experimenting with AI images in 2023 and founded Yuemo Studio, co-produced the AI feature Tai Sui, which went up on Tencent Video in 2025, and later handled all the AI work on the 30-minute film Black Bird, which won a visual award at the Aranya Theatre Festival. She argues that once the tools are easy to use, light, colour, composition and style count for more, not less, and that the field needs hybrid people who understand film, fine art and the models together. Former film teacher Zhang Zefan says companies such as Noon Sunshine and Youhug Media are now hiring for a new job, the "AI producer", which is closer to a pipeline manager in animation than to a traditional producer handling money and people.
Several interviewees say the real gain is flexibility more than cost. Chen says a rough version of a feature can take shape within a month, so directors and investors see characters, pacing and story early. They can then decide where to add real performances or better resources, or stop before the spending mounts, and he estimates about 30% of that early material can survive into the finished film. For investors, the same development budget can test more projects at once.
The limits are spelled out as well. Liu says a script cannot be rushed by adding more generators, and editing, grading, dubbing, effects, music and mixing all remain, sometimes with extra fix-up work to repair drifting characters or mismatched shots. Commercial work needs round after round of generation, selection and revision, so model credits are a real cost: cheaper than a big live-action shoot, but still a serious outlay for an independent creator. Chen Xing, founder of Beijing Tianjiao Star Culture Media, warns about how quickly the technology dates a project: a project should wrap within about two months, because a newer model can make early footage look dated or force it to be redone.
The platforms are building pipelines for this output. Tencent Online Video chairman Sun Zhonghuai said in April that Tencent is making a drama of about a dozen episodes and a 90-minute film entirely with AI. Youku has offered up to 1.35 million yuan per exclusive AI story film since June, paid on viewing results. iQiyi raised the bar for its AI theatre programme from pieces over 15 minutes to features over 60 minutes, picking 10 AIGC commercial films for full production; one is Erosion of Tide by Wang Ze, which is in script, character-design and asset work and is meant to be finished by year's end.
iQiyi's chief executive Gong Yu said in March that cheaper production and more creators will push platforms towards a looser creator ecosystem, and iQiyi has since opened an AIGC Creative Center to individuals, studios and university groups. It offers IP, computing power and funding on tiers from a 10-30% top-up with light review to full financing, where creators must follow the platform's notes. The deeper the platform's stake, the more say it has over script changes, schedule and delivery.
Long Yitong is not worried by the jump in supply. Few teams, she says, can reliably finish medium and long work. She calls the flood of template-built AI comic dramas an "AI Foxconn": text goes in, characters come out of fixed templates, and judgement is left to the model while output grows with more computing power and operators. That will multiply low-quality content, but good long-form work cannot be made that way, just as cameras let everyone take pictures without making everyone a photographer.
She also sees a quiet barrier. A senior platform executive told her most AI creators have ridden a technology windfall and are not yet up to professional film standards, and she believes an AI studio without a traditional team behind it has little chance of getting a feature approved. That is why Yuemo pairs with experienced directors rather than pitching features alone.
The piece closes by comparing the moment to the move to digital cinema, when the Hollywood studios agreed the DCI technical standard in 2002. That shift toppled Kodak and handed lasting roles to tool makers such as Avid, Autodesk and DaVinci Resolve, and the article suggests a similar redistribution of industry positions is now under way. Chen expects a batch of strong AI-made works by the first quarter of next year, and the article notes that the plans of iQiyi, Tencent Video, Youku and Mango TV point the same way.
Sources · 1