CreativeFitting Founder Jiang Zhu: Evolving an AIGC Super Content Platform With the "Trilobite Tactic"
Chinese and American "dominant CEOs" have their differences too.
"ZhenTen" is a rapid-fire Q&A series. Through ten questions, we break down actionable methods for product growth, aiming to surface "non-consensus but correct" insights.
For our first episode, we invited Jiang Zhu, founder and CEO of CreativeFitting, to discuss what a super content platform in the AIGC era might look like.
Jiang Zhu graduated from the Computer Science Department of Shanghai Jiao Tong University. He joined TouchPal, a smartphone startup, after graduation, working across technology, product, business, and marketing — growing from an intern to a partner who rang the bell at the company's IPO. In 2019, he led a team at TouchPal to build a consumer product from scratch: Crazy Reading Novels, which grew from zero to 10 million DAU and 20 million MAU in just a year and a half.
In 2021, Jiang founded CreativeFitting, committed to becoming the world's largest AIGC content platform. That same year, ZhenFund invested in CreativeFitting's angel round. Setting out again, he led his team to land on the short drama box office charts with the AI short drama Five Brothers, while continually searching for the technology and generation paths to quality content.
How will the rise of AI technology reshape short drama creation workflows and consumption habits? How will the world of future imagination expand? Below is the full interview.
Since CreativeFitting's founding in 2021, how would you divide its journey into phases? What was the most critical strategic direction in each phase? What drove each strategic shift?
At the broadest level, we believe a super AIGC content platform will exist in the future, and it should be one close to entertainment.
Initially, we saw the first anchor point for this technology as short video — video-centric content with a certain degree of narrative expression.
We started with a commercially adjacent scenario: short video advertising. Between 2021 and 2022, over roughly a year and a half, we found product-market fit. This was a milestone: one-minute ads with some narrative capability. In this commercial context, many clients partnered with us.
By early 2023, as we continued exploring the technology, we saw the possibility of unlocking continuous, immersive narrative video. So we spent an entire year upgrading in this direction.
In terms of commercial positioning, we judged that short drama would become the next milestone — its commercial value was growing rapidly. So we evolved from short video ads paid for by clients, to short dramas paid for by users.
You've said that DAU isn't CreativeFitting's most critical metric right now. What is your most important goal at this stage?
Our top priority is to produce the highest-quality AI short drama content.
We see AI short drama as our most important milestone today. First, we need to unlock the highest-quality visuals and video content. Second, it needs to generate verifiable commercial revenue. Third, we need to stay open in terms of experience exploration — beyond the video-centric model, we're also building character models so that characters in the drama can interact with users. At the same time, we'll open up some content for user participation, innovating on the experience front.
Over the past year, we've achieved several key breakthroughs.
First, last May, we participated in the Global AI Film Marathon, collaborating with American director Jud Willmont on the AI short film M.A.D, which moved many audiences and won the "Bilibili Audience Choice Award."
Second, we launched Reel.AI, the world's first AI short drama app, achieving initial commercialization. This year, we became the first AI short drama to make it onto global box office charts, generating cumulative revenue in the hundreds of thousands of dollars.
The charting short drama is Five Brothers, an urban romantic comedy created by director Cui Yi, formerly of CCTV, using our AI short drama generation tools.
AI short dramas have long been questioned on whether they can generate user or commercial value. This drama making the global charts has given people hope, validating the scalable commercial value of AI short dramas. AI-produced short dramas can be both critically acclaimed and commercially successful.
If an entertainment-centric super content platform is the ultimate vision for AIGC, what path should we take to get there?
We analogize this startup to biological evolution during the Cambrian explosion. If we go back to that era, in those few short million years, different organisms completed their evolution. Some survived for a very long time and thrived — including the trilobite.
From the start of this venture, we chose a trilobite survival strategy.
Trilobites and Anomalocaris were two organisms that thrived and endured during the Cambrian explosion, but they differed significantly. Anomalocaris took a specialized path, immediately occupying the top ecological niche and preying on lower levels of the food chain.
The trilobite, however, adopted a generalized survival strategy. Simply put: clear strategy, but flexible tactics. Between Anomalocaris and the trilobite, the trilobite ultimately survived longer — over 270 million years.
Like the trilobite, our strategy has three components.
First, build technology ourselves. Deeply understand and keep pace with rapid changes in generative AI technology, especially generative AI-based technology for video and multimodal applications.
Second, stay close to commercial scenarios, because to survive, you need both funding and self-sustaining revenue. Short dramas are growing rapidly both domestically and overseas — possibly one of the few remaining growth markets in recent years. Users have spent tens of billions of RMB, and overseas spending has reached over a billion dollars.
Third, adjust quickly and avoid path dependency. When new insights emerge in the tech landscape, pivot immediately.
Short drama itself is a growth market, and it can combine with advances in multimodal technology. Therefore, AI short drama is a critical path.
How do you define "good" content? How have standards of "good" changed in the AI era for content creation?
Specifically for AI short drama, "good" manifests in two parts: good story and good expression.
Starting with story. Short drama is a highly condensed form that emphasizes dramatic conflict and plot twists. Born in the short video era, it draws heavily from web novel experience but has created an entirely new content form on top of that.
Good expression is equally important. Short dramas are typically distributed on short video social platforms, so they must frontload emotion and conflict, hooking users within 3-5 seconds and delivering dramatic tension within 3-5 minutes, driving continued viewing.
The first step for AI short drama is learning from existing short drama works — especially how to make character performances more natural, with exaggerated dramatic conflict and intense emotion. This has been our key focus in exploration.
But the second step is surpassing today's short dramas. AI short drama is not a replacement for live-action short drama. Its essence lies in AI-generated content that is multimodal — we don't just generate video, we generate characters within the video, and eventually let users participate as characters in the narrative.
Today, when we think about AI content, we must see AI video not just as a projection of an imaginative world, but bring the characters to life and integrate users into it. Users crave immersion in worlds built by imagination for richer entertainment experiences. And as AI and multimodal technology advance, this immersion will only deepen. So I believe AI entertainment, from a user experience perspective, will be very different from today's short dramas.
The "trilobite strategy" means maintaining an open mindset and rapid iteration amid change. What remains unchanged amid all this change?
What doesn't change is our understanding of AI entertainment. This understanding encompasses several dimensions: first, what is the highest level AI entertainment can reach today technologically? Second, what AI entertainment experience do users need, and what is the best expression of that? Third, which AI entertainment applications have higher commercial value?
AI is multimodal, and how different modalities fuse will profoundly shape the final entertainment experience. We believe AI's involvement should continuously increase, but exactly how it should intervene to better serve content expression requires exploration.
We must always maintain respect for the speed at which AI unlocks future possibilities.
Why did Reel.AI launch both AI short dramas and live-action short dramas? What new insights have you gained from observing backend data on consumption behavior for both?
This is actually our gradual evolution path. From the user's perspective, they don't actually care whether a drama is AI-generated. From our perspective, live-action short dramas serve as crucial user cold-start infrastructure and as a benchmark for AI short dramas.
Why were we the first to unlock AI short drama box office revenue? Because we simultaneously understand user needs, understand the commercial benchmarks, and continuously explore AI technology — ultimately reaching this commercial milestone.
The most important insight is this: when we first started creating AI short dramas, users would watch but wouldn't pay. Later, they became willing to both watch and pay.
We realized a critical factor: the transmission of character emotion. One part is facial expression, the other is subtle movements in performance. This is especially important in vertical-screen, female-oriented urban romance short dramas, and it's an area where we spent considerable time achieving technical breakthroughs.
Comparing live-action and AI short dramas makes it easier to identify which critical elements need priority technical breakthroughs. Without the live-action benchmark, it would be difficult to discover how to achieve key commercial breakthroughs.
Here, data serves as the ultimate calibrator for validating conclusions. Whatever hypothesis you propose, only when you see users actually paying with real money do you get clear positive feedback.
But you can't rely solely on data, or you'll miss many open directions worth exploring technologically. We inherently have understanding of content globalization. First, at the story level, we can judge whether a story fits short drama pacing and overseas user needs; at the expression level, we have a feel for content.
This feel can't be deterministically pointed to, but it stems from our long-term accumulated user understanding from novel and gaming products. Relying solely on A/B testing rarely produces true breakthroughs.
From a production standpoint, what is the biggest improvement AI short drama offers compared to live-action short drama?
The biggest variable in AI creation is the ability to serialize — a point that's easily overlooked. Serialization is superior in efficiency, cost, quality, and commercial value.
Previously, when filming a live-action short drama, you had to finish shooting before going live. But AI creation enables a lighter-weight model of updating while producing, making creators more efficient.
Serialization also lets us judge a work's commercial value earlier. Creators can explore their direction faster — whether this creative concept is something users will accept. The serialization model makes what was virtually impossible in the film and television industry possible in the AI world.
In terms of production time, filming a short drama overseas currently takes about one to two months, while our charting short drama Five Brothers took just two weeks.
AI short drama creation resembles writing code — it can be done independently by one person, or through an "open source" model of collaborative creation. So AI creation is fundamentally a more flexible and efficient approach in practice, efficiency, and creative methodology.
What are the aesthetic differences between domestic and international short dramas?
Short dramas definitely can't go global by simply translating scripts. There are significant cultural differences between China and other countries. Take the "dominant CEO" trope: it's popular in both Chinese and American markets, but American female users prefer "double-A" stories — Alpha to Alpha, strong CPs where both male and female leads are powerful.
Chinese female users prefer setups where the male lead is more prominent — the "dominant CEO falls for me" dynamic, where why he falls for her matters less. American women, however, need that process. And South American users favor gangster themes.
Many short drama practitioners come from novel backgrounds. In the global market, many Chinese players have done well with novels — including our past experience at TouchPal with novel products in the US and Southeast Asia — so we have our own understanding of local content culture.
But short drama can't be fully equated to novels. It's a new creation with much innovation to explore. We need to continue exploring around this form; preferences vary across different countries.
Your envisioned endgame for CreativeFitting is an "AI-era super content platform." In this endgame, what would the creator ecosystem and product form look like?
This is an excellent question. We believe an AI entertainment super content platform would first be an entertainment experience where users can watch, chat, and play — where users can't clearly distinguish whether they're watching, playing, or chatting with characters. From the experience perspective, this is certain; but from the creation side, we maintain respect and openness.
Today AI is already generating information — performance scenes, costumes, makeup, props are all AI-generated. But we believe the future should move toward increasing AI's share, with AI participating more in building imaginative worlds, and cross-modal development becoming the trend.
Take Westworld: essentially, robots designed an amusement park where robots have their own lives and entertainment, and humans merely enter this world to complete their tasks and goals. But the park itself doesn't exist for humans — it operates autonomously, and humans merely discover experiences within it.
Traditional video, games, and content are created for humans. Once users stop paying attention, the content ceases to exist.
The AI content world will have autonomous operation capability. It is a world of imagination where people can gain participation, but the world itself does not exist for humans.
What is the relationship between means and ends in AI entertainment? We maintain respect for this question. In future AI entertainment platforms, what will the relationship between humans and AI be? AI is already generating imaginative information; humans and AI are jointly participating in content creation. In certain content categories, AI can achieve 100% generation, even if quality isn't yet optimal, while in other scenarios, human involvement may still be higher. We hope to keep moving forward to see that ultimate endpoint. But today's AI short drama is certainly a very important milestone.
Is building proprietary models necessary to reach the "AI-era super content platform"? What is CreativeFitting's technical implementation path?
From today's perspective, proprietary foundation models aren't necessarily required. Our goal is definitely to build technology ourselves. But building technology ourselves doesn't mean building all technology ourselves. Our team includes world champion algorithmists, but we don't pursue technology as a moat.
If good third-party models exist, we'll incorporate them as part of our stack.
The purpose of technology is to unlock the best AI entertainment experience. We don't pursue a fully proprietary model route, though we are developing our own models. Today we believe unlocking the full AI entertainment and AI short drama content experience requires combining our own technology with others' technology.
When we tried making AI short dramas, we found the industry couldn't yet meet certain technical needs, and we captured these technical gaps with foresight.
We now have an all-in-one AI short drama generation tool. Creators can achieve end-to-end creation on our platform. We focus more on deep understanding of AI short drama, how to use multimodal models to accurately understand creator input, and controllable generation. This tool is essentially our engine.
We use many third-party video foundation models available on the market, but apply them in different scenarios for different purposes. In the generation pipeline, there are many foundational generation plugins. Some we built ourselves, others we call via API. We believe understanding creator intent plus controllable generation is a key element today.
Many directors don't understand AI and can't speak AI's language, but they understand art. So these plugins need to comprehend directors' creative instructions. Controllable generation means that today's short drama creation pursues high quality and high efficiency. But many models or tools aren't designed around these requirements; we need to combine various plugins to maximize generation success rates and efficiency.
Director James Hawes once said, "AI doesn't yet possess human soul and experience." While filming One Life, he worked with Anthony Hopkins. Hopkins always liked staying on set. There was a piano on set, barely related to the plot, but he walked over, lifted the lid, and began improvising beautifully. This moment was ultimately included in the film — such moments are what AI still cannot replicate. This is one of the most beautiful things in creation. Do you believe AI-created content can or will have "soul," now or in the future?
Reel.AI short drama Love Before I Found You Episode 1
This short film was generated by director Wang Pengfei, whose previous work was selected for the Cannes International Film Festival, using our tools. We need to examine so-called "soul" in two parts.
The first part is whether it's a hollow, unremarkable work. I think we need to maintain respect for AI. Today AI's creative capabilities are improving rapidly, and the rate of improvement is exponential. Now it may seem like it's just "close enough," but it will increasingly and rapidly surpass critical thresholds.
The second point is whether AI has consciousness. Let's not define consciousness so broadly — first, does it understand performance? I think AI is slowly beginning to understand performance. Previously, actors were directed to a certain extent by directors, but as AI capabilities improve, its performances will increasingly align with user needs or directors' visions.
Among the top 100 films by box office, over 80% lean toward imaginative content. Of that 80%, roughly 60% is hyper-realistic, like Avatar-style 3D CG blockbusters. This type of content is actually quite scarce — its production share in the overall film market is low, but its box office performance is strong. The problem isn't lack of audience demand, but enormous technical bottlenecks in creating this type of content, with high costs as well.
We need to consider the triangle of quality, efficiency, and cost control. Previously, this imagination-oriented content had high quality, but its efficiency and cost created significant commercial risk. Good works might emerge, but it was hard to produce many of them. But AI lowers the barrier to imagination and shortens the pipeline. We believe there will be a large volume of imagination-oriented works presented in high definition in the future — that feeling will be very different.
Further out, many imagination-oriented content types may not center on traditional novel and film categories, but rather around AI's characteristics. For example, when we shape a character, it may become an IP in the future. We can design personality, story, growth arc — the entire creative approach needs rethinking. So we believe AI will also be a more effective creative approach in the world of future imagination.
The audio version of this episode is also available on ZhenFund's podcast "Seriously Speaking" — tune in!
