Original · Unique Research · 2025-11-17
Editorial note: This is a complete English edition of the historical article and interview. Library size, contracted-author counts, rights ownership, production speed, model capabilities and market assessments reflect the source’s reporting and Zhou Liqiang’s statements, not independently verified current findings. The autonomous multi-agent production scenario is a future vision, not a claim that the complete system is already deployed. The source does not specify a currency for its production-budget range. “COL Digital Publishing” refers here to 中文在线, and “17K Novel Network” renders 17K小说网. Relative time references and forecasts retain their historical context.
As general manager of COL Digital Publishing's AI Animation Division, Zhou Liqiang faces a striking figure: COL has more than 5.6 million (560 × 10,000) digital works and has signed more than 4.5 million (450 × 10,000) authors in total. Its original-content platforms, represented by the 17K Novel Network, continuously produce high-quality content, making it one of China's largest libraries of legitimately licensed content. Many of these works have already proved popular, while some have lower ratings but distinctive subject matter of their own.
For a long time, most of these works had only one form of presentation: reading text.
Only a tiny number of top-tier IP properties had the chance to be adapted into film, television, animation, or games. An industry consensus lay behind this: traditional adaptation is too costly and time-consuming, leaving mid-tier and even long-tail IP with little chance of making the queue.
Before AI emerged, there was almost no solution to this reality. Now generative AI has made the question worth reopening: can more works be turned into visible content in less time and at lower cost?
Can the opportunity for adaptation be extended, as far as possible, from a handful of leading properties to a much wider range of works?
COL Digital Publishing is working by breaking this question down step by step.
I. AI Animated Short Dramas: Giving More IP a Chance to Prove Itself
The company's current focus is using novel IP to generate AI animated short dramas.
In one sentence: it turns a full-length novel, through a highly automated production line, into an animated short-drama series that can run on both long-form and short-form video platforms.
This approach primarily addresses two longstanding problems:
First, production cost and production cycles.
Traditional animation, film, or television can easily take months or even years, with budgets ranging from hundreds of thousands to tens of millions.
Under that model, only a very small number of top-tier projects can be undertaken.
With AI, a work can move from script and storyboard through art and final footage on a timeline measured in days, while costs also come down.
That gives a chance to works that previously would never even have been considered for adaptation.
Second, content supply cannot keep pace with consumption.
In the short-video era, the amount of content users scroll through each day is astonishing, while conventional production capacity struggles to match that pace.
The advantage of AI animated short dramas is that the mindset can shift from producing one work to producing an initial batch and then examining the data.
An IP property does not have to begin with an immediate bet on dozens of episodes. A team can first produce a few episodes as a trial and use real data to decide whether to invest more resources.
Under this model, the order of IP development is rearranged. Adapting an IP used to be one major, high-stakes decision; now it is more like launching on a small scale first and letting the data speak.
Some works may demonstrate their potential within a few episodes; those that perform only moderately can be stopped before further losses.
For platforms, authors, and production teams alike, this is far more practical than spending years only to discover that a project does not work.
At the same time, animated short dramas also feed audiences back to the original work.
Many users first encounter the short drama and then go looking for the original book.
Stories once encountered only by people willing to sit down and read long-form text can now reach a much wider audience first through short dramas.
II. An AI Production Workflow, Not a Single Model
Judging only by the technical terminology, it is easy to assume that projects like this merely connect several large models.
What COL Digital Publishing is actually building is closer to a production-management system: AI is used to reconstruct the entire filmmaking process.
The company calls it an AI producer workflow, and it includes several key roles:
First, the AI screenwriter.
A proprietary large model reads full-length novels, understands character relationships and story structure, and then works with human editors to divide the novel into episodes, establish the pacing, and create storyboards.
AI handles the rough processing, while people make judgments and fine adjustments to ensure that the adapted story remains coherent and its central conflicts stay clear.
Second, AI art.
Vertical style models are trained for different genres. Chinese fantasy and cultivation have one aesthetic; contemporary feel-good romance has another; suspense and science fiction follow different systems again. AI's task here is to generate character sheets and concept art for settings with a consistent style and stable character appearances.
Across every episode, the protagonist's face and each important setting must be recognizable at a glance, so viewers know who they are seeing and where the scene takes place.
Third, AI animation generation.
The script + characters + settings are used as inputs, and a video-generation model handles camera movement, performance, and duration control.
COL Digital Publishing is not seeking live-action-style realism at all costs. What matters is continuity and emotional expression within the language of animation: whether a character's movements and facial expressions match the dialogue and the progression of the story.
People have not been pushed out of this chain; instead, they occupy a new position:
They spend more time on aesthetic judgment, deciding what must change, and controlling the overall rhythm, rather than repeatedly polishing every frame.
AI performs large volumes of repetitive work; people decide what can be used, what must be redone, and what requires manual refinement.
This change in the division of labor deserves more attention than model parameters.
It shows that when AI is truly deployed, what changes is not one technical point but the way the entire team works.
III. From Workflow to Agent: Applications Will Become Systems for Automated Collaboration
The next step brings us to Agentic AI.
Today's process still requires a person to act as dispatcher:
Once the screenwriter finishes, a person checks the result before passing it to the art team;
after the art is produced, a person confirms whether it fits the genre and market positioning before sending it to video production;
each stage is connected to the next by a person.
Introducing Agentic AI would automate this process further.
A more ideal future state would look like this:
The team would need to issue only one relatively complete task instruction, for example:
Adapt this novel into a 100-episode animated short drama for primary distribution in North America, with a style similar to a specified work.
A producer Agent would then launch the entire process:
automatically call a screenwriting Agent to perform multiple rounds of adaptation,
call an art Agent to complete character and setting design,
call an animation Agent to generate each episode,
call a voice Agent to produce localized audio,
and even call a distribution Agent to create accounts on target platforms, publish the content, and analyze the data.
The entire process could continuously check and correct itself.
People would focus mainly on two things:
setting a clear objective and judging whether the final result deserves to be scaled up.
The form of the application would consequently change in an obvious way:
it would no longer be merely a tool, but an automatically collaborating system.
The difference between companies would also shift from which models they have connected to how they organize and manage an entire digital workflow.
For COL Digital Publishing, this evolution naturally leads to managing the full IP life cycle.
That spans content selection, automated incubation, and distribution across multiple markets and channels.
IV. Taking AI+IP Global: Multiple Languages and Regions, Not One Template for the Whole World
When it comes to globalization, COL Digital Publishing's advantage can be summed up in one word: content.
Chinese online literature has performed well overseas in recent years. Genres such as Chinese fantasy, cultivation, and feel-good romance have established audiences in many regions.
Here, AI animated short dramas further lower the barrier to understanding, allowing users who find reading Chinese difficult to enter a story quickly through visuals and plot.
With AI dubbing, multiple language versions can be generated simultaneously. The cost of simultaneous distribution in English, Spanish, Arabic, and other languages is incomparable with the past.
Technically, the reach of the content is significantly amplified.
But globalization is not simply translation + multilingual publication. It also involves three difficult issues: cultural differences, compliance standards, and values.
Many plotlines that are highly popular in China may not be understood in other regions;
the complexity of xianxia settings and certain conventions of emotional expression are received differently across cultures.
At a deeper level, different regions have different thresholds for violence, gender issues, and religious content.
COL Digital Publishing's approach is to combine AI with local teams:
AI handles high-efficiency production, while local teams judge whether the content is appropriate and how it should be adjusted.
In priority markets such as North America, Southeast Asia, and the Middle East, the company works with local content-operations or advisory teams.
Their job is to conduct the final content review: checking whether dialogue flows naturally, whether character settings could easily be misunderstood, and whether any story beat crosses a red line.
COL Digital Publishing also primarily handles IP for which it owns the rights, rather than user-uploaded content, so its data-privacy and compliance pressures are relatively manageable.
Even so, as countries around the world continue to strengthen regulation of AI-generated content, this remains an area that requires long-term investment.
V. Chinese AI Companies: From Application Innovation to Content Brands
If the global AI industry were laid out on a map, the position of Chinese companies would be relatively clear:
They are still catching up in foundation models, but they stand out in application-layer innovation and commercialization speed.
The Chinese market is characterized by a large user base, rapid feedback, and fierce competition.
Once a new approach proves effective, it is quickly scaled—and quickly attracts new competitors.
In this environment, knowing how to use AI to build a viable business often matters more than simply leading on technical metrics.
Zhou Liqiang believes Chinese AI companies must address two gaps in the future:
First, they must move from point applications to long-term ecosystems.
They cannot stop at building an App that is popular for a while or capturing one wave of traffic.
They must consider where their real long-term advantage lies.
For COL Digital Publishing, that advantage is IP + AI: it has both a content library and the ability to use AI to make incubation more efficient.
Only by building a complete content ecosystem around those two strengths can it create a relatively durable moat.
Second, they must move from taking products overseas to building brands overseas.
In the past, many companies considered the job complete once they pushed a product onto overseas rankings.
In the future, the more important question will be:
When overseas users think of a particular kind of AI content, will a Chinese brand naturally come to mind?
For example, AI animated short dramas made by Chinese teams might become known for dense storytelling, distinctive world-building, and fast pacing.
Once that association is established, it benefits every product that follows.
This step is difficult and takes time, but without it, the advantage will remain limited to cost and speed and struggle to move to a higher level.
VI. Individuals and Small Teams: More Tools, and More Responsibility
The theme "Pioneering Intelligence | The Individual Era" is repeated so often because it points directly to a very real change:
The spread of AI tools really does make it possible for one person to do work that once required an entire team.
But that does not mean the barriers have truly disappeared. They have moved.
The old barriers were whether someone could draw, edit video, or create 3D work.
Today's barriers lie more in three areas:
First, the ability to collaborate with AI.
Instead of personally completing every task, people must learn how to make requests effectively, select among results, and iterate continually.
AI can give you many options at once, but choosing the final one and deciding how to adjust it requires clear judgment.
Second, understanding and controlling the workflow.
Do not place the entire bet on a single tool.
Tools will be replaced and interfaces will change, but the path from idea to finished work is relatively stable.
Individuals and small teams should invest more effort in whether their own process—from inspiration to publication and then review—works smoothly, rather than memorizing the operating details of one piece of software.
Third, a change in role.
Many creators are accustomed to thinking, "I have something I want to express."
In an AI-enabled content environment, however, a creator must also assume part of the role of product manager and producer:
understand who the target users are and what they care about;
design the pacing, genre, and length so that viewers are more likely to finish the content;
and decide how much to spend on an initial experiment and under what circumstances further investment is worthwhile.
AI lowers the barriers at the execution layer and enables more people to begin;
but the ability to keep going is still determined by these less glamorous capabilities.
COL Digital Publishing's experience shows that:
AI can help a content library containing more than 5.6 million (560 × 10,000) works improve efficiency and give more stories a chance to be adapted and seen;
the same logic applies to individuals and small teams.
How many ideas you have is not the key question. What matters is:
whether you are willing to break them into steps that AI can execute,
and whether you are willing to use data and feedback to decide at which version to stop.
Tools will proliferate, and interfaces will become increasingly convenient.
In the end, the works that will be remembered over the long term are those polished with care, along with the people behind them who are willing to keep adjusting their methods.
Selected Interview Q&A
Q1: Please briefly introduce yourself and COL Digital Publishing.
Zhou Liqiang: I am Zhou Liqiang, and I currently lead COL Digital Publishing's AI Animation Division. Since COL was founded in 2000, its work has always revolved around two keywords: content and IP. We have more than 5.6 million (560 × 10,000) items of legitimately licensed digital content, making ours one of China's largest content libraries. From PC reading to mobile reading, audiobooks, film and television, games, and animation, we have always been thinking about one thing: how to help a good story reach more people.
Q2: What is the core focus of your AI work today?
Zhou Liqiang: Our flagship effort is "using novel IP to generate AI animated short dramas." Put simply, we are building an automated production line from text to visuals so that a full-length novel can appear on long-form and short-form video platforms as an animated short-drama series. The goal is not to make a flashy demo, but to give more IP a chance to be adapted and validated.
Q3: What industry pain points does this product primarily address?
Zhou Liqiang: The two most practical ones are these. First, traditional animation, film, and television take too long and cost too much, which means a vast amount of mid-tier IP never receives an adaptation opportunity. Second, content is consumed so quickly in the short-video era that traditional production capacity cannot keep up. AI compresses the production cycle from months to days and also lowers costs substantially. We can first conduct a small-scale trial and use the data to decide whether to invest more.
Q4: If you had to state your original aspiration in one sentence, what would it be?
Zhou Liqiang: Do not let good stories remain buried in the library.
We have too many excellent works that exist only as text, which is a pity. Generative AI gives us a chance to present the scenes in our imaginations more efficiently and at greater scale, allowing more people in more countries to see them. That is the starting point of our "AI for IP" approach.
Q5: Which technical breakthrough in generative AI today is most important to your business?
Zhou Liqiang: The combination of video generation, multimodal understanding, and controllability.
The new generation of models can do more than produce clips. They can understand physical laws and cinematic language and preserve the consistency of characters and settings over time. For our animated dramas, this determines whether shots connect coherently and whether characters keep unexpectedly "changing faces."
Q6: How do you turn these frontier technologies into products?
Zhou Liqiang: We approach it through the idea of an "AI producer." An AI screenwriter reads novels and adapts scripts, AI art creates characters and settings in a consistent style, and an AI animator handles camera movement and performance. AI completes most of the work, while people focus more on judgment and fine-tuning to ensure that the pacing, aesthetics, and emotional tone remain strong.
Q7: With technology evolving so rapidly, how do you balance "innovation" and "execution"?
Zhou Liqiang: Our principle is straightforward: applications first, technology in support, and IP at the core.
We do not necessarily need to build the underlying large models ourselves, but we must be able to command them effectively. The technical team focuses more on fine-tuning, workflows, and efficiency. Commercially, we adhere to the MVP approach: use AI to produce a launchable version quickly, let the data speak, and then decide whether to refine and scale it or stop before further losses.
Q8: How do you understand Agentic AI? Will you use it?
Zhou Liqiang: We are watching it closely and exploring it internally. For us, Agentic AI is the ultimate form of the "AI producer." People are still needed to connect the stages in today's workflow. In the ideal future, I would issue only one task, and the producer Agent would coordinate the screenwriting Agent, art Agent, animation Agent, voice Agent, and distribution Agent automatically, then give me the result and the data. This changes the form of an application—from a single tool into a digital team.
Q9: What new opportunities do AI animated short dramas create for taking Chinese content overseas?
Zhou Liqiang: Chinese online literature already has an audience abroad. Genres such as Chinese fantasy, cultivation, and "domineering CEO" romance have fans in many regions. Previously, reaching them depended mainly on translation, which created a relatively high barrier. AI animated short dramas now offer a more intuitive way to reach new users. Combined with multilingual AI dubbing, content can be released simultaneously in English, Spanish, Arabic, and other languages, creating an entirely different level of efficiency in going global.
Q10: What is the hardest part of globalization?
Zhou Liqiang: First is cultural discount: settings that Chinese audiences find highly satisfying may not be understood by or resonate with overseas users. Second, different regions have different review and ethical standards for religion, violence, and gender issues. Our approach is "AI + local people": AI provides production capacity, while local teams handle the final 20% of localization by revising dialogue, adjusting character settings, and avoiding red lines.
Q11: What role do you think Chinese AI companies play on the global stage?
Zhou Liqiang: At present, we are the most agile application innovators and the largest real-world testing ground.
We may still be catching up in foundation models, but Chinese teams iterate extremely quickly when it comes to "using AI to build a real business." The next step is to move from individual applications to ecosystem building. For example, we hope to turn "IP + AI" into an integrated system spanning content discovery through global distribution, while moving from "taking products overseas" to "building brands overseas."
Q12: How do you interpret the theme "Pioneering Intelligence | The Individual Era"?
Zhou Liqiang: For me, its essence is that AI becomes an amplifier for the individual.
In the past, producing an animation required a company, funding, and time. Now, one person with an AI workflow has a chance to make one. AI sharply lowers execution costs, which makes creativity, aesthetic judgment, and decision-making even more important—the one-person company is no longer a slogan but a very practical way of working.
Q13: For creators making AI animated short dramas, what are the greatest opportunity and challenge today?
Zhou Liqiang: The greatest opportunity is the benefit of a paradigm that is still taking shape. No fixed playbook has emerged, and major companies have not yet completely monopolized the field. Anyone with strong aesthetic judgment and a keen sense of online culture may be able to use one workflow or one prompt to create a hit. The greatest challenges are homogenization and the "tool trap": because everyone uses similar models, content can easily acquire the same generic flavor. Tools also change so quickly that people may spend all their energy chasing them instead of refining the story itself.
Q14: What advice would you give individuals or small teams? How should they position themselves over the next 1–3 years?
Zhou Liqiang: Three points:
First, develop the ability to collaborate with AI. Put your energy into asking good questions, selecting outputs, and exercising aesthetic judgment, rather than competing with AI over "who can draw better";
second, build your own dynamic workflow instead of betting on one tool;
and third, move beyond being solely a creator toward the role of "product manager + producer," learning to design content from the perspective of the market and audience. In the Individual Era, you are both a creator and your own CEO.