Original · Unique Research · 2026-04-13
Editor’s note: The first-person report and its judgments belong to the original Chinese author. This English rendition retains the opening essay, repeated examples and full panel, including all 25 named speaking turns and their continuation paragraphs. Revenue, salary, user, output, customer, share-buyback, investment, efficiency and market figures are source or speaker claims, not independently audited findings. Monetary figures without a currency in the original—including the headline’s 100 million—remain without an inferred currency; views are not revenue. The source is dated April 13, 2026, but alternates between “next year” and “2026” for the 50% AI-imagery forecast. Both are preserved, without inventing an event date or treating the prediction as achieved. Statements that Sora opened up and was not open to consumers, and that Midjourney was restricted from going abroad, retain the source’s conflicting or unclear wording rather than establish product availability or regulation. The named Stanford/Fei-Fei Li affiliations, company/customer relationships and image-production totals are attributed, not independently established here. The Minecraft ‘nearly US$2 billion’ figure and its World Model characterization remain the speaker’s statement; the translation does not reinterpret the figure as cumulative revenue. Zhong’s roster title, Senior Product Engineer, differs from his self-description as senior creative lead and algorithm/data engineer; both remain. Company, personal and work titles are transliterated where official English forms remain unverified. ‘Backward training’ retains the source’s technical wording rather than guessing a different method. The investor ‘slap in the face’ quotation is figurative rhetoric. The opening account says all buyback funds were earned from products; the later interview also acknowledges an early financing round. These source formulations are retained without inferring a cap table or audited financing history.
Unique Awards
After 13 Years, He Bought Back All the Shares in His Company
In an age when AI is reshaping content production, what actually keeps a tools company alive may not be the most cutting-edge model terminology, but whether it has found a relatively unchanging need.
"
The best way to convince investors is to give them a slap in the face with data and money.
Yan Huapei said he had not met with investors for a long time.
This was at a Hangzhou AI WEEK roundtable. Five people deeply involved in content-production tools were sitting together when someone happened to ask him: what do you do if investors define a tools company as SaaS and you cannot raise money?
He did not mince words: “The best way to convince investors is to give them a slap in the face with data and money.”
He had grounds for saying it: he had been an entrepreneur for 13 years and had bought back all outside shareholdings along the way. Every bit of that money had been earned through the products.
I think this is a key to understanding the roundtable. The people there were not discussing AI’s possibilities; they were discussing how to use AI to actually survive on a specific battlefield.
Five Different Answers to the Central Question
Moderator Duan Hongyu opened with a question: in content production, does AI create more value through efficiency or through creativity?
The five people gave five answers, none exactly the same.
Chen Weiyu is China Literature’s head of AI, overseeing the paths by which major IP such as Douluo Dalu and Doupo Cangqiong are being brought into AI. His judgment was the most straightforward: “The biggest improvement right now is still in efficiency. Creativity itself remains a uniquely human skill.”
Liang Wei used to make films. Now he works on MovieFlow, turning AIGC tools into a workstation that film and television professionals can use. He was more optimistic: “Many ideas that could not previously be realized will be brought to life as these tools become widespread. Creativity will certainly be stimulated without limit.”
Yan Huapei said efficiency and creativity reinforce each other: “Enormous improvements in efficiency are also helping creators extend their creativity without limit.”
Zhao Lu comes from Xufeng Technology, which has Fei-Fei Li’s team behind it and works on spatial intelligence and world models. He offered a phrase I thought was more precise: “the manual labor within intellectual work.” A designer has an idea but must spend a great deal of time on rigging, revisions and rendering. This is the manual labor that knowledge workers have to do, and it is precisely what AI frees them from.
Zhong Xiangyu works on Meshy, one of the earliest companies in the global 3D-generation field to expand internationally. His formulation was more direct: “It is not an efficiency gain, but the evolution of an ability from nothing into existence.” Through AI and 3D printing, ordinary people have, for the first time, gained the ability to turn what is in their minds into something tangible they can hold.
After hearing all five answers, I noticed something: they were not really arguing about efficiency versus creativity. They were discussing what AI had made possible that had previously been impossible.
“3 People, 5 Days, a 100-Minute Narrative Feature”
Liang Wei gave a figure. MovieFlow’s Studio version was still in internal testing, but their test result was that a narrative feature of around 100 minutes could be completed by just 2–3 people in 5 days.
What was the process like before? He said, “It took several months to go from handing in the screenplay to preparing for filming.”
The logic has now changed: first spend 5 days making the film once with AI, then assess which parts are worth shooting live and which do not need to be shot. Essentially, this brings the cost of trial and error down to a minimum.
Liang Wei said the largest film companies, television-series companies and platforms in China were all testing this version internally.
Then he made a prediction: by next year, AI imagery should account for around 50% of traditional film and television.
In 2026, half the imagery would be AI-generated: this was not distant science fiction, but the trend he saw in the internal-test data.
I am rather curious to see what the Studio version’s actual output looks like. Put “3 people, 5 days, 100 minutes” in front of any conventional Chinese film crew and they would treat it as a joke—but there the data are.
He also said something else that I think has been somewhat underestimated:
“The smallest unit of AIGC creation is one director + one screenwriter + one editor. Those three people cannot be replaced.”
Not a technical director, not a producer, not marketing. Director + screenwriter + editor. Those are the three components he believes cannot ultimately be removed from content creation.
China Literature’s Answer: Even Tenfold Screenwriter Pay May Not Be the Ceiling
China Literature’s AI comic-drama business passed 100 million last year, producing nearly a thousand titles, of which 12 exceeded 100 million views. By March this year, the number of works with more than 100 million views had reached 26, a rapid increase. This was the result of its “IP+AI” strategy.
The underlying logic has three layers: IP is the source of good stories, screenwriters are the core of creativity, and AI accelerates production.
Chen Weiyu described a very specific application: the comic-drama assistant can quickly read through an online novel of ten million Chinese characters, extract the emotional storylines in fiction for female readers, and give screenwriters inspiration for their story setups. Previously, a screenwriter needed a week to read a book all the way through; now AI can do it.
But he immediately followed that with a statement pointing in quite the opposite direction:
“With AI helping screenwriters free themselves from trivial, low-value work, they can devote themselves wholeheartedly to creating the story itself, and their future opportunities will be greater.”
Why? He said tools cannot solve the problem of coming up with good ideas. That is a task for AGI, and the current stage is not there yet. Producing assets with AI is becoming easier and easier, but combining them into a hit story that moves people and rises above homogeneous competition still requires genuinely creative people.
The more abundant the supply, the scarcer good screenwriters become. The logic is counterintuitive, but on reflection, it holds up.
“We Want to Find the People the Giants Overlook and Small Teams Cannot Serve”
Shanjian has been going for 13 years. At its core is a gap in the market: users whom large teams overlook and small teams cannot serve. They are called KOC—specifically, salespeople who need to acquire customers through platforms, such as real-estate agents, insurance advisers and beauty-product salespeople. They need to keep producing content but lack the time and ability to make videos.
Yan Huapei described what they want to achieve in a short phrase: “lazy to the extreme.” Say anything, or forward a link, and it automatically generates a formatted Xiaohongshu image-and-text post or copy for WeChat Moments. There is no need to understand prompts or the logic of the tools. It simply works.
Yan Huapei said billions of people around the world still had not benefited from even very traditional offerings. Sora being released or Midjourney being restricted—those fluctuations at the frontier did not have much direct bearing on his users.
“Technology changes so quickly that even short dramas are now called ‘traditional short dramas.’ So entrepreneurs have to find something relatively unchanging to feel grounded. Otherwise, they are anxious every day.”
I think this is useful for many people trying to find their direction in the AI wave. Everyone is chasing the newest models and capabilities, but many parts of users’ needs are extremely stable: there will always be people who need to turn ideas into content and who need tools to save them time.
Those needs will not disappear because GPT-5 or Sora arrives. They will only become more pressing.
“People Using 3D Printing Have Gained the Ability to Turn Imagination into Physical Objects for the First Time”
Meshy works on 3D generation, with customers including major overseas companies such as Supercell, SEGA and Snap. In games, it has improved overall efficiency at the concept-design stage by 2–3 times, cutting an art asset’s completion time from 10 days to 1.5 days or even within a day.
Zhong Xiangyu said 3D printing had now become one of the core engines of growth for the entire AI 3D industry.
Why are people using 3D printing more willing to pay than game companies?
For these people, AI offers not just efficiency, but an ability that did not previously exist. An ordinary person could not previously turn something in their mind into a 3D model on a screen, let alone print it and hold it. Now they can.
“It is not an efficiency gain, but the evolution of an ability from nothing into existence.”
That statement explains a great deal. For professionals, AI is an efficiency tool; for ordinary people, it is a new way of existing. Both markets are large, but the psychological mechanisms that drive payment are completely different.
What Someone from Fei-Fei Li’s Team Said
Xufeng Technology has a Stanford-incubated research background in Fei-Fei Li’s laboratory, with a focus on spatial intelligence and world models. Its domestic applications cover three areas: real estate and home renovation broadly; culture and tourism, including work with a location-based VR company in which director Zhang Yimou is involved; and service robots.
Zhao Lu gave a figure: humans had produced more than 15 billion images over the past 200-plus years, but the images generated by AI in the last three years had already exceeded 20 billion.
The purpose of the figures was not to show how impressive AI is. He immediately added that AI has enormous potential to stimulate creativity.
Then he used the phrase I found most precise: “the manual labor within intellectual work.”
A designer has a good plan but needs three days to implement it in Blender. A screenwriter has a good structure but needs a week to read through the original work and find the emotional arc. These are manual tasks that knowledge workers have to perform. Having AI free up that part is not laziness; it genuinely leaves people time for the 20% of work that can produce creative ideas.
He also discussed their approach overseas: in controllable generation, they had not followed Midjourney’s “excessively creative” route, because industrial-scale collaboration with businesses requires controllable output, not randomly generated surprises.
There is a tradeoff between creative freedom and controllability. A little more creativity may mean a little less value for an enterprise customer. That tradeoff is everywhere in B2B settings.
Find What Does Not Change
After listening to the roundtable, I noticed something interesting: the least anxious person there was the one who had been an entrepreneur the longest.
Yan Huapei had been doing this for 13 years and had lived through too many moments of “What do we do now that a disruptive technology has arrived?” His answer was simple: find a relatively unchanging need among users, then serve it to the utmost.
Sora arrives; Midjourney is restricted. These waves may come every few months. But people’s desire to produce content is always there.
I kept thinking about his words: “Entrepreneurs have to find something relatively unchanging to feel grounded.”
This is not advice to give up following the frontier. It means that, before following the frontier, you need to know what problem you are trying to solve.
If you know the problem, you can use any tool that comes along. If you do not, even the most advanced tools just offer another way to get lost.
More Details from the Conversation
Unique Awards · Hangzhou AI WEEK Trends Roundtable Panel
“Creation Is No Longer a Test of Stamina: How AI Tools Reshape Content Production’s ‘Efficiency Curve’”
Guests:
China Literature — Deputy General Manager of Technology & Head of AI — Chen Weiyu
MovieFlow — Co-founder — Liang Wei
Shanjian Intelligence — Founder — Yan Huapei
Xufeng Technology — President, China — Zhao Lu
MeshyAI — Senior Product Engineer — Zhong Xiangyu
Moderator: Unique Research — Partner — Duan Hongyu
Duan Hongyu: I am Duan Hongyu, the moderator of this Panel. This Panel will mainly discuss how AI improves the efficiency curve in content production. Yesterday, everyone saw two news items flood WeChat Moments: one was that Sora had opened up, and the other was that Midjourney had been restricted from going abroad. Because everyone here has brought products in content generation, we will have a brief discussion of these topics with our guests today. First, please take about a minute each to introduce yourselves and your companies’ businesses. As for how AI improves content production, the market generally sees two major roles: improving the efficiency of content production and improving creative capabilities. Please give your view on whether AI’s current capabilities play a greater role in efficiency or creativity.
Chen Weiyu: I am Chen Weiyu from China Literature. You can call me Dayu. I am mainly responsible for our AI business, including the comic-drama tools that are popular now, as well as internal AI tools such as our copyright assistant. China Literature’s domestic business is primarily in copyright. We have a vast collection of copyrighted IP, including Douluo Dalu and Doupo Cangqiong, and we also have many offline exchanges and collaborations.
As far as AI tools themselves are concerned, I think the biggest improvement right now is still in efficiency. I still consider creativity itself a uniquely human skill. Good content that meets the standard for release still requires human creators, or highly creative individuals such as the star writers on our platform, to be involved.
Liang Wei: Hello everyone, I am Liang Wei from MovieFlow. Sorry, my voice is rather nasal because of allergies. We are primarily a global AIGC tool, currently available both in China and abroad, covering more than 170 countries with around 1 million users. Our backend has been live for less than half a year and has already generated over 5 million minutes of video. We are now introducing the Studio version, both to enable professional film and television creators to make AIGC feature-length works in the future and to genuinely enable project management, cost reduction and efficiency gains with AI across the traditional film and television workflow. It is also a very good asset-management center for companies generating text, images and video with AI.
The moderator just raised the question of efficiency versus creativity in AI video capabilities. Because I used to make films, I believe the traditional film and television industry will definitely move actively into AI. Traditional IP, copyright and directors all need to be rebuilt in the AIGC world; everything is worth doing again. Many ideas that could not previously be realized will be brought to life as tools become widespread. Without restrictions on content review and distribution, creativity will certainly be stimulated without limit.
In terms of efficiency, in internal testing a couple of days ago, we found that a narrative feature of around 100 minutes could essentially be completed in five days by just two or three people in our Studio version. That used to be unimaginable. Previously, it took months to get from handing in the screenplay to preparing for filming. Now you can first spend five days making the film once with AI, then assess which live-action parts need shooting and which do not. Efficiency is greatly improved. Although Sora has not opened to consumers and is in high demand, models keep emerging and will not stop. We very much hope everyone will embrace this world more.
Duan Hongyu: Mr. Yan, I remember that you also came from film and television.
Yan Huapei: Yes, coincidentally, I used to make films too. I worked on films at Enlight in ’06 and ’07—looking back, that was already 20 years ago. Our company has actually always made video tools, for both China and international markets. Shanjian is just one of our products, centered on digital humans. We are now trying to turn digital humans into what we call ‘Digital Human 2.0’: not merely lip-syncing, but storing a person’s full set of data and media assets, then using that asset library to help automatically create short videos, image-and-text posts, articles, podcasts and WeChat Moments posts.
Shanjian currently primarily serves the KOC group—for example, salespeople who need to acquire customers through platforms. Recently, we have been working with many retail brands that have thousands or tens of thousands of salespeople, hoping to become the central platform that empowers their sales teams.
To answer the earlier question—efficiency or creativity—I think they reinforce each other. If I suddenly have an idea but cannot realize it immediately, improvements in execution efficiency allow that idea to keep building on itself and change along the way. When we made films back then, a ten-second shot took a long time to model and render. Today, a child might say something offhand and the idea comes to life. Enormous improvements in efficiency are also helping creators extend their creativity without limit. It is a wonderful era for people with ideas. A lot of poor content will flood in too, but that is normal at the beginning.
Zhao Lu: I had a brief discussion with the moderator offstage earlier. I am Zhao Lu, head of Xufeng Technology in China. We are a research-background team incubated at Stanford in Silicon Valley. Everyone should know the laboratory’s best-known Chinese scientist: Fei-Fei Li. Our direction is also spatial intelligence and world models. Our overseas team is called Collov AI. We have nearly 7 million registered users overseas, covering almost 70 countries.
In China, since last year, we have covered three commercialization scenarios:
Real estate and home renovation broadly: generating spatial worlds from images.
Culture and tourism projects: we are working with some local governments and with a location-based VR/AR company in which director Zhang Yimou is involved.
Service robots: because we work on spatial training and world models, we were already exploring open- and closed-source approaches and publishing papers in ’21 and ’22. There will be considerably more collaboration with robotics companies this year.
I come from a planning and design background. In design, much of our time is spent on ‘the manual labor within intellectual work.’ Once an idea emerges, you still spend a great deal of time doing rigging and revisions with tools such as Blender or Maya. AI frees creativity and means people no longer have to spend so much time on this manual component of intellectual work. There is a balance here. In controllable generation overseas, we did not follow Midjourney’s excessively creative approach, because that makes industrial-scale collaboration with companies such as Philips difficult. It depends on each team’s vertical use cases and definition of commercialization.
Duan Hongyu: So when customers pay for your product now, are they more interested in the creativity it produces or in its efficiency?
Zhao Lu: If creativity is the greater focus, the improvement in creative efficiency is definitely more than 20-fold. But if customers place particular emphasis on Control and being Scalable, some efficiency has to be sacrificed, and the increase may be only a fewfold.
Zhong Xiangyu: I am Meshy AI’s senior creative lead, and I also work as an algorithm and data engineer for large models. Our main business is 2D and 3D generation, serving games, film and television, and 3D printing. Our customers are mainly overseas. We may be the earliest company in 3D generation to have expanded internationally. We began with 3D meshes, then gradually expanded into 3D printing and turning creative ideas into physical deliverables.
As for whether AI can improve efficiency or quality, I very much agree with what Mr. Yan said. Take games: the industry went through a period when large numbers of reskinned games driven by paid user acquisition burst onto the scene. That was an inevitable phase. AI generation will inevitably bring a lot of low-quality content too, but ultimately it will raise the ceiling on creativity and quality. So Meshy is currently focused primarily on improving efficiency, while the next priority is improving creativity and quality.
Duan Hongyu: From everyone’s explanations, your fields are quite different. For example, China Literature announced last year’s short-drama results in mid-March, and revenue was very good. I recently spoke with people making dramas in Hengdian. Everyone in live-action short dramas is now making premium work; almost nobody is making poor-quality short dramas, and platforms are largely moving toward premium content too. The comic-drama market is now also experiencing severe sameness and poor quality because of mass generation with AI. How does China Literature use AI to build an advantage in your field? Could you describe the central value it provides?
Chen Weiyu: On how to define a premium story, as someone who has enjoyed online fiction for twenty or thirty years, I would say a good story first needs a foundation: a plot that moves people. That is the IP. Second, in comic dramas, you need a new setup that breaks through existing sameness. Third is production. Everyone extracts assets from large models, so why do they look different in some people’s hands? That comes down to production techniques. These three aspects make a good story.
To help good stories emerge, China Literature has made the following arrangements:
Story sources: make the rights to good IP that readers have already ‘voted for with their feet’ over China Literature’s history available to partners.
Story setups: provide the comic-drama assistant. Previously, a screenwriter needed a week to read an online novel of ten million Chinese characters. Now AI can quickly read the whole book, extract—for example—the emotional storylines in fiction for female readers, and inspire the screenwriter’s setups.
Production: customized styles. The style of the images determines that of the video. Through Lora or backward training, we adapt large models to produce a distinctive visual style, such as characters from a particular period of the Tang dynasty.
Duan Hongyu: Quite a few people attending today run traditional short-drama businesses, and many small and medium-sized companies are still watching and waiting when it comes to comic dramas. On one hand, Token costs are not particularly low. On the other, comic-drama screenwriter salaries in Beijing have increased tenfold. Could your AI comic-drama tools lower the requirements placed on people somewhat and bring costs down?
Chen Weiyu: First, comic dramas cost far less than short dramas. Low cost is also the key to their explosive growth: it is what enables efficient ROI and commercial returns. Computing costs are actually relatively manageable. Comic dramas and short dramas do have two different audiences. Short dramas may cater to older users in lower-tier markets, while comic dramas appeal to younger users. Both have room in the market.
As for labor costs, creativity will become increasingly valuable as AI tools spread. That creativity comes from screenwriters, so their future opportunities will be greater. Even a tenfold increase may not be the ceiling. Tools cannot solve the problem of producing good ideas; solving that would be a job for AGI. Many people can use tools to produce assets, but screenwriters and directors still need to steer them into an extraordinarily creative hit story.
Duan Hongyu: Mr. Liang, MovieFlow serves film and television post-production professionals. How do you improve their efficiency in your use case?
Liang Wei: Our Studio version is currently in internal testing. China’s largest film companies, television-series companies and platforms are all testing it. Going forward, traditional film and television and AI-generated imagery will coexist. By next year, we expect AI imagery to account for around 50% of traditional film and television.
MovieFlow’s Studio version is designed entirely around how film and television professionals work. A crew has more than a dozen specialties: directors, producers, costumes, makeup, props, cinematography, art departments and so on. When faced with AI, they do not know where to start. They cannot simply pick up a tool and reduce costs or improve efficiency, so we provide a workstation that genuinely serves film and television.
Consumer AIGC content will emerge across the board this year. As traditional film and television companies, directors and screenwriters enter, they will greatly increase the value of IP. You can remake Journey to the West and Water Margin with AI, and people around the world can remake their own religious and cultural stories. From 2026 onward, content formats will become more professional. People will genuinely pay for AIGC, and this era’s House of Cards—or outstanding films and long-form series—will certainly emerge.
Duan Hongyu: One more question for Mr. Liang. Last year, China’s short-drama market was a little over a hundred billion, while traditional film-industry revenue was around fifty billion. In other words, short dramas were roughly twice the size of film. Does MovieFlow have plans for short dramas or comic dramas?
Liang Wei: Our Studio workstation can actually be used for either short dramas or comic dramas. Of course, it will be easier for traditional film and television professionals and fit their working logic better: preparation after the screenplay is finished, asset building, shooting scene by scene, then continuity and collaborative creation.
I have always thought the smallest unit of AIGC creation is ‘one director + one screenwriter + one editor.’ Those three people cannot be replaced; ideally, they also have producing capabilities. Professional film companies such as Wanda are now entering one after another. The AIGC era serves global, digital content, a market far, far larger than the hundred-billion-scale short-drama market we see today. People’s love of stories will not change. Film box office will not grow or shrink because of short dramas; it is a normal, ongoing reality. People need to get used to traditional film and television coexisting with AIGC entertainment.
Duan Hongyu: Mr. Yan, we know Shanjian is an extraordinary company. It achieved a hundred million in revenue very early on without a sales team. You just said you are now moving toward applications for salespeople. How do you improve efficiency through the product, and how do you address sameness?
Yan Huapei: Our company has actually been going for 13 years. We raised funding once in the early days, and a couple of years ago we bought back all the shares. So from the beginning, we had to find a way to earn enough to support the team. Technology changes very quickly; even short dramas are now called ‘traditional short dramas.’ So entrepreneurs have to find something relatively unchanging to feel grounded. Otherwise, they are anxious every day.
Many people worry about what we will do now that Sora has arrived. I think that although we are discussing the frontier here, billions of people around the world still have not benefited from even very traditional offerings. There are plenty of giants making editing tools, such as Jimeng, Kling, Pika and Runway. The only way to survive among them is to find a highly specific group of users and serve them deeply, however the technology changes.
Creative people will need different tools in the future. We used to find it magical to animate wind and water in Maya; people exchanging prompts today are doing much the same thing. Efficiency has simply gone from a hundredfold to a thousandfold or ten-thousandfold, but people’s desire for magical images has not changed. Shanjian hopes to serve deeply that group in between whom small teams cannot handle and giants overlook. We want to make users ‘lazy to the extreme’: say anything or forward a link, and it automatically generates a formatted Xiaohongshu image-and-text post or a WeChat Moments post, enabling them to enjoy the value of content to the full. Find a group to serve deeply, and you can last longer without the anxiety.
Duan Hongyu: A related but slightly off-topic question: as financial advisers, or FA, when raising money for video-tool companies, we often encounter investors defining them as SaaS companies. Without pursuing cutting-edge technology concepts, a company labeled SaaS may be unable to raise money. What do you think?
Yan Huapei: I actually have not met with investors for a long time. Back when our first tool, Doupai, reached number one, all the investors came. Early on, entrepreneurs need data and results to prove themselves. Getting paid is a result. Whether it is called SaaS or something else is just a concept to make communication easier.
Do not use old ideas to define business models in a new era. The best way to convince investors is to ‘give them a slap in the face with data and money.’ For entrepreneurs building tools, the scariest thing is not investors disbelieving your story, but becoming ever more deeply stuck on the wrong path yourself. Opus Clip, for example, was not originally in this field either; it proved its value through user numbers and data. It is better to step outside the labels and genuinely explain the value to users.
Duan Hongyu: Mr. Zhao, you serve real estate, culture and tourism overseas. Could you elaborate on the main practical scenarios in which you currently improve efficiency?
Zhao Lu: I strongly agree with Mr. Chen and Mr. Yan about focusing on users. As a technology-driven team, we have commercialized successive generations of products and become number one by penetrating a particular niche. A product may not have a moat, but a specific user segment will certainly use it. Early Minecraft, for example, was effectively an early version of a World Model, and it sold for nearly US$2 billion. Many children were crazy about it while adults could not understand it. That shows that different eras will inevitably have different groups using new tools.
We develop within the Vision Pro ecosystem. In the culture and tourism market, for instance, AI can upgrade content that previously cost several million to produce but offered very poor immersion. Humans produced more than 15 billion images over the past 200-plus years, but AI-generated images in the last three years have already exceeded 20 billion. AI has enormous potential to stimulate creativity.
Duan Hongyu: Mr. Zhong, similarly, where specifically do the efficiency gains appear in your 3D field?
Zhong Xiangyu: Meshy’s clearest practical application is in concept design and prototyping for games. In the traditional workflow, when major overseas companies develop the concepts for a game world, teams fly abroad to take well over a hundred thousand photographs, then go through complex modeling to turn them into 3D assets. The process is extremely long. Game companies using Meshy currently see overall efficiency gains of about two to three times in concept design. Many art-outsourcing companies can, through multiple rounds of client review, cut an asset’s turnaround from 10 days to 1.5 days or even within a day.
Besides games, another focus is 3D printing, which is also a priority for our international strategy. Users of 3D printing may already account for more than half our revenue. Individual users need to turn what is in their minds into images or 3D models on a screen, and ultimately into physical objects they can hold. With AI and mature 3D-printing technology, ordinary people have gained the ability to turn ideas into physical things. This is no longer simply an efficiency gain, but the evolution of an ability from nothing into existence.