Just now, a 33-year-old man (born in 1992) secured a financing of 926 million yuan.

Hey, today a future unicorn has emerged in the AI application track: a 92-born young man has raised $130 million (926 million yuan) in just two and a half years of entrepreneurship.

What kind of company can raise so much money? The answer is LiblibAI, an AI company specializing in image generation. More specifically, its business consists of an image generation tool and a community.

The investors in this round include Sequoia Capital China and CMC Capital. Existing shareholders such as Shunwei Capital and Source Code Capital have also increased their investments. Comparing domestic investment and financing data, this is the largest single financing in the domestic AI application field so far this year.

ByteDance is truly the cradle of AI entrepreneurship in China. The founder of LiblibAI is Chen Mian, who was previously in charge of commercialization for the Jianying and CapCut teams. Congratulations to the alumni of Southeast University for cultivating another future unicorn for the industry.

Through the financing event of LiblibAI, we can analyze several new opportunities in the industry.

1. 3D content generation.

As the demand for 3D content in industries such as gaming and robotics continues to rise, 3D AIGC technology is becoming a new hotspot.

2. Instant content generation.

For example, the generation time of high-resolution images can be reduced to the millisecond level.

3. Deepening of vertical application scenarios, such as e-commerce, healthcare, and education.

In short, there are still plenty of new opportunities in the AI image generation track. New players, go for it!

01

The boss of LiblibAI is a post-90s, born in 1992, and was once an employee of Zhang Yiming.

After graduating from university, he joined ByteDance and was in charge of commercialization for the Jianying and CapCut teams. He was one of the youngest "4-1" (middle and senior management) product managers at ByteDance.

At the end of 2022, ChatGPT set off an AI boom. Chen Mian observed ByteDance's tool product line and identified a trend: AI is starting to change the way of creation, but "those who understand AI can't create, and those who can create can't use AI well."

If AI can't be used by more people, how can the technological dividends be released?

In May 2023, he left ByteDance to found LiblibAI with the goal of creating an AI product that allows every ordinary person to easily create. "Creation is human nature, but the threshold was too high in the past. The significance of AI is not to make designers unemployed, but to enable more people to express themselves," said Chen Mian.

In the early days of entrepreneurship, the LiblibAI team had only a dozen people, mostly from Internet companies such as ByteDance, Meituan, and Tencent.

Its first product was an AI image generator, but Chen Mian soon realized that it was difficult to build a moat based solely on technological output.

So, the team decided to upgrade the product to a "creator community" to form a differentiated ecosystem through model sharing, work display, and community interaction.

In the summer of 2023, LiblibAI became a bit popular through user self - promotion, with the number of users exceeding one million in three months. However, at the end of the same year, the AI track cooled down, user growth slowed down, and the platform once faced an operational crisis.

"At that time, server costs, computing power costs, and community maintenance were all burning money. We once thought the company might not survive," recalled Chen Mian. To stay afloat, the team began to build its own model system, optimize computing power utilization, and explore commercialization through membership and creator revenue - sharing models.

In early 2024, LiblibAI's creator function was launched, allowing users to upload their own LoRA models and set them for paid use. The platform transformed from a tool - based product into an AI creator ecosystem.

In the same year, the company completed multiple rounds of financing, with a cumulative amount of hundreds of millions of yuan. The team expanded to nearly a hundred people, and the number of users exceeded ten million.

In early 2025, the number of monthly active creators on the platform exceeded three million, and the cumulative number of generated content exceeded 500 million images, making it one of the largest AIGC creation communities in China.

02

Next, I'll focus on its products. In general, compared with general large models, it is a vertical model + community.

In terms of functionality, its core product is an "AI canvas" - users can input text, upload images, or set styles, and the AI can generate high - quality image works. The platform supports users to train their own models, share materials, and build a creation community.

Currently, LiblibAI has gathered millions of active users, most of whom are designers, e - commerce practitioners, illustrators, brand operators, and independent content creators. They have generated hundreds of millions of images and tens of thousands of models on the platform for various scenarios such as product promotion, film storyboarding, illustration design, and game concept art.

The industry where LiblibAI belongs is: large models - vertical large models - large models for image generation. Of course, in addition to it, large models for text and video are also very popular, with different application scenarios.

The main application scenario of LiblibAI is drawing. This field has roughly gone through three stages:

• Germination period (around 2021): AI drawing was still confined to the scientific research or open - source circle, with complex models and high thresholds.

• Explosion period (2022 - 2023): Overseas tools such as Stable Diffusion and Midjourney became popular, and AI creation began to enter the public eye.

• Localization and ecosystem period (since 2024): Domestic platforms have gradually emerged. The focus is no longer on "whether it can generate", but on "whether it is easy to use, fun, and can generate revenue".

Looking closely at the pain points in this track, the main issue is that although traditional design tools are mature, they cannot meet the real - time requirements of AI - generated content. In simple terms, they are too slow, inefficient, and rely too much on human labor.

LiblibAI's solution is: model tools + community. Indeed, if there were only model tools, a startup would have limited competitiveness when facing giants. With a community added, once the product is chosen by a large number of users, traffic migration and usage habits will become LiblibAI's moat.

However, it's not easy to build a successful community. The value of a community lies in its network effect - users come for the content and contribute content, which in turn attracts more users.

But this effect is very fragile.

First of all, to grow the community, continuous large - scale traffic acquisition is a problem. In today's era of saturated Internet traffic, pure traffic purchase is not only expensive but also inaccurate. The cost of attracting an AIGC creator is much higher than that of attracting an ordinary app user.

This is a weakness for startups but a strength for giants (such as ByteDance). Giants can use the large user base of their mature products for "bundling" or "traffic diversion" (for example, embedding community entrances in their cloud platforms or office software), while startups have to start from scratch and acquire users one by one.

Secondly, to strengthen the community, the more difficult problem is how to retain the traffic. A user may download your tool just to use a certain function. How can you make him realize that the "neighboring" community is an indispensable part of his workflow? This requires excellent product design to deeply integrate the community into the core process of the tool.

Overall, the core capabilities of a community are very different from those of model tools. Whether the LiblibAI team can meet the requirements of both is a question that only time can answer.

03

Looking back at the entire image AIGC track, the domestic market is currently small but growing at a good pace.

According to industry estimates, the market size in 2024 was close to 20 billion yuan, and the compound annual growth rate is expected to exceed 30% in the next five years.

However, competition is intensifying. Many domestic AI creation platforms have emerged, such as Huimengdao, Huizhi AI, Krea, PixVerse, etc. Some focus on image generation, while others specialize in video or virtual humans.

Although LiblibAI doesn't position itself as a "generation tool" but as a "creation infrastructure", the actual difference isn't that significant, and there is direct competition at least in terms of functionality and application scenarios.

Additionally, this track also faces "cross - border competition". Why? Because AI creation is entering the era of "multi - modal integration". Put simply, when working, users need to generate text, images, and videos - it would be best if one tool could have all three functions, which would best meet user needs.

Therefore, the competitive landscape of the future image AIGC track is still highly uncertain: How powerful can AI become? Will the ultimate product form be multi - modal integration or single - modal deepening? We really don't know.

But according to the trend, there are several opportunities worth considering for new players.

1. 3D content generation. Currently, judging from public information, LiblibAI's users mainly generate planar content. As the demand for 3D content in industries such as gaming and robotics continues to rise, 3D AIGC technology is becoming a new hotspot.

2. Instant content generation. AIGC technology will break through in real - time generation capabilities, aiming to shorten the generation time of high - resolution images to milliseconds.

3. Deepening of vertical application scenarios, such as e - commerce, healthcare, and education. In the medical field, AIGC can be used to generate visual materials of various pathological features to assist medical research and teaching; in the education field, it can be used to reconstruct ancient scenes and character images, making abstract knowledge more intuitive. In the next 1 - 3 years, the demand for AIGC image production in these fields may further increase.

The development of AI applications is far from reaching its ceiling. New players, go ahead!

This article is from the WeChat official account "Pencil News" (ID: pencilnews). Author: Truth - teller. Republished by 36Kr with permission.