What is Liblib’s moat?
Article author and source: 36Kr
Recently, the AI video space secured another major funding round. Evoken, Liblib’s parent company, has completed a B+ round of approximately $300 million co-led by Granite Asia, Tencent, and Shunwei Capital. The company has a post-money valuation of over $2 billion, making it a new unicorn in the AI applications space.
Image source: Weibo screenshot
This means that after Aishi Technology, another company in the AI video generation space has reached the $300 million funding milestone. However, unlike Aishi Technology, Yanyu Technology is a product company that has grown rapidly by aggregating mainstream models available on the market for designers and creators without developing their own models.
As large model iterations accelerate, a single model update can overwrite most of a product’s functionality. A hot topic in the industry is whether the AI application layer still has a real moat. Despite this background, why did Yanyu Technology attract the support of capital despite the trend? And is the product really competitive enough to withstand model iteration?
When discussing a company’s ability to attract capital, the focus is always on the founder. Chen Mian, founder of Yanyu Technology, has worked at several leading technology companies, from Tencent, 360, Baidu, Didi to ByteDance. Before launching the startup, Chen served as global head of monetization for CapCut and Jianying at ByteDance. Born in 1992, he became ByteDance’s youngest Level 4-1 employee at the time.
ByteDance’s background may have made Chen Mian more attractive to investors. Tianyancha said the company secured an angel round of funding from Source Code Capital, Gaoyu Capital, and Jinsha Ventures just two months after its founding. As of June this year, Yanyu Technology has completed six funding rounds, with Tencent, Ant Group, Sequoia Capital, and Shunwei Capital participating in the recent B+ round.
Image source: Tianyancha screenshot
Beyond Chen Mian’s career, Yanyu Technology’s rapid growth in terms of spending is supported by its user base and revenue trends. Yanyu Technology’s core products include Liblib AI, an AI creation community launched in 2023. Xingliu and Lovart (overseas version), an AI design agency launched in 2025. and LibTV, an AI video creation platform launched in March of this year. Among them, Lilibib AI has accumulated over 30 million users, making it one of the largest AI asset websites and creator communities in China.
Based on the disclosed ARR, the company’s total annual revenue is over $300 million. Five months after launch, Lovart achieved over $80 million in ARR. Although specific numbers regarding LibTV were not provided, Yanyu Technology said that LibTV’s monthly revenue increased more than 13 times within two months of its launch.
Lovart gained great popularity overseas by addressing the pain points of fragmented design tools. Traditional AI design tools only generate individual images. After users enter the prompts and receive the image, they must manually switch to Photoshop for editing and Figma for layout. Lovart allows users to generate complete design solutions at once.
The same goes for Liblib. Aggregate multiple models and provide video creators with a more streamlined workflow. This convenience is the basis for the success of tool-based agents.
Image source: Screenshot from LibLibAI official website
“Qu Jie Shang Ye” noticed that Yanyu Technology was rapidly iterating its products. When LibTV was first released, some users new to AI editing said they couldn’t learn fast enough to keep up with software updates.
Unlike Lovart, LibTV was fortunate to launch at the perfect time, coinciding with the industry’s transition from live-action to animated shorts. After this year’s Lunar New Year, Hongguo Short Films actively entered the AI short film field, prompting many short film producers to seek reliable video generation tools. To date, LibTV has already gained a significant user base among short film creators, who currently represent a core consumer group in the video generation field. Additionally, LibTV’s key users include teams that need to produce visual content and marketing materials at high frequency, such as the creative departments of top advertising agencies and brands.
However, LibTV’s biggest competitive advantage when compared to Streamlined Workflow is its price.
Due to a lack of computing power and a surge in users, Mind4 adjusted its prices three times in April, increasing video production costs nearly six times. As a result, LibTV, which also provides access to “Seedance 2.0,” is becoming increasingly popular as a cost-effective alternative for many AI video professionals.
According to LibTV’s official website, LibTV’s creative members are currently divided into different levels: standard, advanced, and premium, each with different video generation limits, with fees ranging from 569 to 8,499 yuan per year. Some editors have pointed out that the team has been gradually switching video production to LibTV since Ji Meng raised the price in April. There will be fewer editing tutorials and it will be harder to find answers to questions, but low cost and low latency are the most important factors.
Source: LibTV screenshot
Some comic creators point out that LibTV can generate one minute of content for as little as 20 yuan. It is very difficult to lower Ji Meng’s prices, and such low prices may only be possible under strict conditions such as off-peak computing power and limited time.
As a “model migrater”, why can Liblib offer lower prices than the original platform?
This touches on the commercial nature of aggregation platforms. Liblib provides authors with access through an API interface from model providers. All computing power consumed by Liblib comes from the model provider’s computing center. In other words, user spending on Lilibib primarily depends on the cost of purchasing computing power, and Lilibib earns a margin as an intermediary.
If Liblib commits to consuming a certain amount of computing power per token each year, the discount rate could be lower. However, this discount is not so significant that LibTV can sell it to users at a much lower price than Ji Meng and still make a profit.
To remain “cost-effective” on the C-side, there are only two options: subsidize users with cash or look for cheaper channels such as intermediate APIs to obtain compute power allocations.
This may also be the reason why Yanyu Technology raises funds frequently. Without the ability to develop models, they rely primarily on subsidies to retain users. This is very similar to the early cash battles of mobile internet, where platforms used subsidies to scale and cultivate user habits.
However, if creators choose Liblib because of Dream’s price increases, they may abandon Liblib if a cheaper platform emerges or if Dream lowers its prices.
It is also worth noting that platforms often overlook content compliance during the stage of gaining market share. In April of this year, China Central Television Financial reported that LilibAI was experiencing an “AI-generated explicit content” issue, allowing users to circumvent moderation by generating prohibited content using subtle prompts. LilibAI has since apologized and said the technical fix has been completed.
It is essential to strengthen the prevention of security risks and compliance issues with AI tools.
Image source: Weibo screenshot
An avid hard-tech investor bluntly stated that the moat for AI aggregation platforms is shallow. These platforms are easily replaced as larger companies open APIs, lower prices, or release better native applications. More importantly, compared to the strength and speed of current model iterations, the differentiating advantages of these aggregation platforms are very fragile, and features refined over six months can become obsolete in a single model update.
This is not just a Liblib issue. Most AI application layer tools face similar structural challenges, as their product value is highly dependent on the functionality of upstream models.
Whether AI will kill software has also been a recurring topic of discussion this year.
In January of this year, Anthropic released Claude’s tool Cowork and industry-specific plugins covering areas such as legal, finance, and sales services, replacing some vertical SaaS software and causing multiple drops in Nasdaq Software stock. Anthropic CEO Dario Amodei has previously expressed similar views. As large-scale models mature, software functionality such as video generation and graphic design will be incorporated into the generic underlying model, eliminating the need to develop and sell third-party tools on your own.
However, NVIDIA CEO Jensen Huang publicly pushed back against the notion that “AI will exhaust software,” arguing that agents will be built on top of enterprise systems and structured data. Software providers need to transform, not replace.
The relationship between software and AI is not yet agreed upon within the industry. As a result, some characterize platforms like Lilibib as “intermediate products before the model converges.” LibTV achieved more than 13x increase in monthly revenue within its window of opportunity, making it a rare winner because “fast” is itself a rare ability.
However, Lilibib’s continued growth will depend heavily on continued funding and its ability to evolve from a “model replicator” to a must-have creative infrastructure for users before the window of opportunity closes.
