th working with Hou Zuxin, a Chinese director who had never used AI before, to make an AI-generated short film. "Can we find a bridge … and see if they are not perceived as AI movies, but as movies?" Wachholz told CNA. It is a question increasingly facing China’s screen industry, where AI is moving from the margins to the mainstream of content production - at every scale, from 90-second vertical dramas to feature films bound for cinemas. CNA Games Guess Word Crack the word, one row at a time Buzzword Create words using the given letters Mini Sudoku Tiny puzzle, mighty brain teaser Mini Crossword Small grid, big challenge Word Search Spot as many words as you can Show More Show Less But as AI reshapes how films and television are made, it is also disrupting jobs, changing the skills filmmakers need and fuelling debate over whether it can ever replicate the creativity and emotional depth of human storytelling. One glimpse of that transformation came at the Shanghai International Film Festival (SIFF) in late June. This year, the festival introduced an AI Backlot programme, pairing traditional filmmakers with AI creators to produce an AI-generated short film over the course of a month. Wachholz and Hou were among the participants. Across China's screen industry, AI is increasingly becoming part of everyday production. Some of the country's biggest entertainment companies have publicly recognised the rapid advances AI is making in the industry. Gong Yu, the CEO of streaming giant iQiyi, predicted that purely AI-generated feature-length films could appear as soon as this summer and that, within five years, more than half of leading film and television works could be AI-generated, according to his interview with Chinese business publication China Entrepreneur, published in May. Tencent Video’s chairman announced in April that the company had completed a high-quality, long-form AI-generated drama. According to local news outlet Yicai, he framed AI as a force in the screen industry, reorganising the content supply chain by allowing small creative teams to produce work that once required large crews, while intensifying competition and raising the bar for standout productions. That same month, the National Film Administration licensed China's first AI-generated feature for cinemas - a 90-minute science-fiction film from Bona Film Group, classified as animation. But nowhere has AI's impact been more dramatic than in microdramas - the low-cost, vertically shot serials watched billions of times by Chinese audiences. The microdrama market overtook China's annual box office for the first time in 2024 and is forecast to top 120 billion yuan (US$16.5 billion) this year. About 128,000 microdramas were released in China in the first quarter of 2026 - nearly four times the total a year earlier. More than 95 per cent of them, or about 122,000 titles, were AI-generated, according to an April report from the China Netcasting Services Association. That is, titles generated entirely with AI, without cameras or live actors, rather than live-action productions that use AI as a production tool. A year earlier, the figure was near zero. For supporters, the appeal of AI is straightforward: It promises to make films and TV shows faster and far cheaper to produce. Using AI, a team of five to 10 people can now deliver a full microdrama series in two weeks to a month, a fraction of the time and manpower required for a comparable live-action production. One AI-generated

microdrama series costs around 100,000 yuan all-in and goes from script approval to launch in under a month, according to Junjun, a producer at a Beijing-based company that has pivoted from live-action titles to exclusively producing AI-generated microdramas. Junjun asked to be identified only by that name and for her company not to be named as she was not authorised to speak to the media. By comparison, a traditional live-action costume microdrama series costs at least 300,000 to 500,000 yuan and takes two to three months to make, according to Chinese news reports. For Wang Ze, a director trained at the Shanghai Vancouver Film School, AI has changed what was possible. Selected for an AI filmmaking showcase at SIFF, he made a seven-minute animated short in about 10 days - a project that would have required a much longer production schedule and a crew he did not have if made conventionally. "In the past, to make a work, it took cost, it took manpower," he told CNA. "Now I can borrow AI tools and make the film I want all by myself." But Wang also pointed out that the bar for the creator using AI is raised higher, describing a director who now also needs a working grasp of cinematography, storyboarding, lighting and art direction. For Wachholz, the Berlin filmmaker, AI’s key appeal is access. Describing film as “gatekept on so many levels”, he said that AI lowers the barriers to entry. "People who can't draw but have good ideas can create images." Hou, the Chinese director who partnered with Wachholz at SIFF, still works primarily in live-action filmmaking. She still shoots features, documentaries and commercials, and sees AI not as a replacement for filmmaking, but as another production tool. It can fill gaps that budgets or schedules cannot, she said. For example, generating an establishing shot of a coastline that a client cannot afford to travel to film. Learning to direct an AI model is simply another language to master, Hou said, no different from communicating with an actor, a colourist, or a sound designer. "It's just more tools we can choose,” she told CNA. "The most important thing is to know what you want.” What unites the filmmakers is the belief that AI can transform how films are made without replacing human creativity. Veteran director Huang Jianxin, who runs Xiamen University's film school and chairs the Beijing Film Association, distilled that view into a simple rule. "The stronger the tool, the more important human judgment becomes,” he said at SIFF. "Don't deify it, don't demonise it, don't fear it.” Yet the efficiency AI promises comes at a cost. As productions require fewer people, editors, extras and other crew members whose work can be automated are among the first to feel the impact. Chinese outlets have documented the downturn, which they attribute to AI exacerbating an already oversupplied market. A survey by tech-commerce outlet Ebrun found more than 60 per cent of live-action crews in the microdrama sector had quietly halted work after the Spring Festival - concluding that AI was "not the direct killer, but the last straw", atop a glut of titles and the scrapping of platform subsidies. Financial newspaper 21st Century Business Herald reported mid-tier actors' fees halving and backstage roles - lighting, costume, photography - shedding jobs in large numbers. Shanghai-based news outlet The Paper reported in May that crews which once posted casting calls now advertise for AI roles instead. Junjun, the producer,

described the labour shift at ground level. An editing team that once needed five or six people now needs one, she said, because the AI generates video with part of the cut already built in. Some editors became "generation artists", or choukashi - a gacha-gaming term for generating batch after batch of footage and keeping the usable frames, she said. The company's contracted actors have been hit harder, she added. “There really are fewer roles they can take now". Actor Bi Zhiqiang has felt the impact firsthand. He has worked in Hengdian, China's drama production hub, since moving there from Shandong in 2021. Two years ago, work was plentiful. "As long as I wanted to shoot, I could shoot non-stop," he told CNA. "Sometimes on a single day, many crews would send me job offers, but I could only accept one." The offers dwindled around Chinese New Year this year, leaving him with idle days he had rarely known. Now, he said, a production crew calls once or twice a month. The drop in demand coincided with the surge in AI-generated dramas. Bi declined to share his earnings before and after the slowdown. Sun Bin, vice-dean of the School of Drama, Film and Television at the Communication University of China, said AI has permanently changed how dramas are made - a rebuilding of the industry's production model, rather than a “short-term traffic bubble" of hype that will deflate. Sun named the first jobs to feel the impact. "Hengdian extras, basic lighting, cheap set crew, and pure editing operators,” he said. Creative roles, by contrast, are likely to become more valuable, Sun said, with directors, writers and other jobs requiring artistic judgment less exposed than entry-level production work. For all its rapid advances, industry insiders said AI still struggles with some of filmmaking's most fundamental challenges. Wu Hankun, an actor and content creator, has experimented with using AI to re-render real, filmed performances - and believes something is lost in the conversion. "Some details of a real performance simply can't be reproduced." Junjun, the producer, has encountered similar limitations. In her experience, AI has "very little sense of spatial orientation", meaning shots can lose visual continuity and characters drift out of position. Vertical format hides much of it, she said. With less of the scene on screen, a character’s face can draw a viewer’s attention away from a background that does not quite hold. Sun from the Communication University of China said that while AI is already ahead at creating fantastical worlds and producing content at scale, it still struggles to match the subtleties of human expression. "AI has no life, no joy or anger or sorrow," he said. "All its emotions are data imitation." Bi, the actor, expressed similar sentiments. "AI might know what falling in love means, but it's never been in love,” he said. Bi also does not see AI replacing his profession entirely. "AI is capable of replacing a portion of actors … for higher-skilled actors, it'll be hard to replace them in the short term,” he said. Bi intends to stay in his line of work, even as friends and crew members drift to other avenues. Han Lichen, a director at a studio that produces wholly AI-generated content, said the flood of AI shorts is "assembly-line stuff, with no core or substance". "AI can only help us generate good-looking visuals," he told CNA. "The core lies in the human." Xie Danming, iQiyi's senior vice president and chief AI scientist,

illustrated the challenge with a beach scene in which a girl runs, pauses and turns back. AI can generate each shot convincingly, he said, but when they are stitched together, subtle inconsistencies in the weather, lighting or even the movement of the waves can make the sequence feel unnatural. iQiyi stated in March that while AI can make a 30-second video, "creating long-form narratives at professional drama and film quality remains an industry challenge". Cost remains another obstacle, despite AI's reputation for lowering production costs. Wang, the director, said polishing AI footage to the point where the seams do not show can mean "doubling the compute", referring to the amount of processing power required. The extra computing requirements also drive up costs, meaning work good enough for professional standards “is actually not as low-cost as everyone assumes", he said. Asked where the screen industry might be in five years, most interviewees did not expect today’s wave of wholly AI-generated titles, particularly in the microdrama segment, to define the future of filmmaking. Instead, they expect AI to become increasingly embedded in conventional productions, while fully AI-generated content remains concentrated in lower-cost, high-volume formats. iQiyi’s Xie described two distinct modes already in practice. In genres that rely heavily on digital production - including animation, children's content, manga adaptation - entire productions can be generated by AI, with the human team acting as what he called an "AI director" overseeing the script, artistic vision and final edit. For large-scale live-action dramas, by contrast, AI serves mainly as an advanced visual-effects tool. “Human teams, live-action shooting and cameras remain the absolute core" of these projects at present, Xie said. More than 95 per cent of new microdramas released in China may now be AI-made, but Sun from the Communication University of China said that headline figure masks how much of the output is disposable. The bulk of it, he said, is a “transitional chaos" of low-cost, formulaic work that he expects tighter regulation and rising audience taste to clear out. Even viewing figures come with caveats. Without giving numbers, Junjun, the producer, said her company’s AI-generated offerings currently attract slightly larger audiences than live-action titles. But she did not put it down to quality. Much of it, she said, is driven by novelty. People click because the work is AI and they are curious about what the technology can do. Audiences are also more forgiving of AI's flaws than they would be of mistakes made by human filmmakers, she argued. At the same time, the quality of some live-action microdramas has fallen low enough that their AI-generated counterparts need only clear a relatively low bar. "It might just have hit a baseline level, and you already find it more interesting than an ordinary video,” Junjun said. Sun acknowledged that technology may transform the way films are produced. "But the core of film art will always be people, story, emotion, value,” he said.

微软披露AI收入大部分来自OpenAI

据彭博社报道，微软最新披露的文件显示，在截至6月的财年中，微软从OpenAI获得241亿美元收入，占其AI业务实际收入的约70%。微软CEO纳德拉此前表示，截至3月底公司年化AI收入已超370亿美元。尽管微软试图通过投资Anthropic和自研模型来减少对OpenAI的依赖，但投资者仍质疑其AI收入中OpenAI的占比。分析人士指出，这一披露可能与OpenAI的IPO计划有关。目前尚不清楚OpenAI的贡献中有多少来自收入分成协议，多少来自云计算等服务。 #微软 #OpenAI #AI收入 #科技新闻 #财报 #人工智能 #IPO