From Images to Video, Meitu Has Found a More Lucrative Business
The explosive growth of video applications has been one of the defining phenomena in the AI industry over the past year.
From Higgsfield, Flova, and LibTV to Adobe’s Firefly and Meitu’s KePai and MVLAND, startups and leading players in the visual content space are attracting market attention—or undergoing revaluation—through video creation tools and platforms.
These video applications have several things in common: they orchestrate model capabilities into workflows and build creative platforms and content ecosystems through Agents, infinite canvases, Skills, and other formats, helping users produce better results more efficiently. This has renewed market confidence in the application layer, while also signaling that video applications are entering an intense process of elimination.
Having entered the video application market, how to build a defensive moat through distinctive product strategies and execution has become a new challenge for players in the visual content space.
It’s Not About the Model, but the Workflow: The Boom in AI Video Applications
With 2026 more than halfway over, the narrative direction of the AI industry has shifted once again.
As model capabilities continue to evolve and converge, the market has gradually reached a consensus: users care about the result, not where it comes from. It does not matter which model generates a video. What matters is efficiency, quality, and monetization potential.
The reality behind this is that “models swallowing applications” has not happened. Instead, the relationship between the model and application layers is gradually moving toward collaboration. In other words, models provide the underlying technology, while applications provide platforms and workflows that deliver more stable and consistent results to users. In terms of business models, model companies make money by selling computing power, while application companies make money by selling outcomes.
Of course, general-purpose model providers may still enter the market directly and compete with application companies. But in the first half of 2026, the surge in AI visual applications—particularly AI video tools—was the first to unlock the application layer’s potential. These applications orchestrate model capabilities into workflows, then use Agents, infinite canvases, Skills, and other tools to build AI content production systems, transforming their products into creative platforms that can deliver results directly to users.
In the primary market, video creation tools such as Higgsfield, Flova, and LibTV have attracted considerable interest. In August, AI image and video generation platform Higgsfield completed a $400 million Series B round; in June, Yanyu Technology, the parent company of LibTV, completed a nearly $300 million Series B+ round.
In the public markets, attention to these video applications has focused more on leading players in the imaging sector. Adobe’s Firefly, for example, is an “all-in-one AI creative workspace” designed for creators, covering images, video, audio, design, and Agent workflows. Among Chinese companies, Meitu has entered the video application market with products including Wink, KePai, Vmake Labs, and MVLAND.
More Vertical, More Focused, and More In-Depth
Unlike Higgsfield and Firefly, Meitu’s AI video applications represent a different product philosophy and approach: more vertical, more focused, and more in-depth.
This approach appeared to have been established as early as the beginning of the AI boom. Meitu launched its first video product, Wink, in 2022. It quickly gained traction through highly vertical features such as image-quality enhancement and video retouching.
With the launch of KePai and Vmake Labs in 2023, Meitu truly entered the productivity application market for video creation. KePai and Vmake Labs target the domestic and overseas markets, respectively, focusing on talking-head video and marketing video workflows.
In China, KePai allows users to create talking-head videos efficiently through features such as a teleprompter, one-click video production, and internet-savvy templates. It helps users in high-value industries such as insurance, education and training, and health and wellness acquire customers online, quickly becoming the leading product in the talking-head video segment. In the United States, Vmake Labs has likewise gained popularity in the fitness and health and wellness markets through talking-head marketing videos.
At the 2026 Meitu Image Festival, KePai launched the “KePai AI Assistant.” Through a 24/7 AI talking-head video team composed of multiple Agents, it covers the entire process from “account analysis, content selection, and script planning to video generation and editing,” continuously delivering high-quality talking-head videos to users.
On the one hand, this product approach aligns closely with the current trend of video applications focusing on workflow orchestration and creating commercial value for users. On the other hand, it is more concentrated on vertical scenarios and industry users. Combined with Meitu’s more than a decade of accumulated expertise in aesthetics and industry know-how, integrating these capabilities into repeatable workflows enables greater precision, consistency, and controllability in the results.
What is often overlooked is that Meitu has another, deeper moat compared with general-purpose model providers or application startups: its understanding of aesthetic preferences, detailed user needs, high-quality results, and industry trends, continuously accumulated through users’ generation, editing, saving, and frequent use of content. Meitu can then use these insights to inform the fine-tuning of its proprietary vertical models and product iteration.
Put more simply, in the types of vertical scenarios targeted by KePai, it is difficult for any other product to understand what users want better than KePai does.
The commercial performance of KePai and Vmake Labs also demonstrates the viability of this approach. According to Meitu’s latest financial report, as of June 2026, KePai’s MAU had grown by more than 100% year over year, its number of paid subscribers had grown by more than 150%, and its ARR had grown by more than 100%. Meanwhile, Vmake Labs’ ARR reached approximately $5 million.
Specialization Is Becoming a Moat
Anish Acharya, a partner at a16z, recently said in an interview that specialization is becoming a new type of moat. His explanation was that today’s technology is already capable of enabling product teams to become highly specialized for a very specific user group. In other words, an understanding of users, the ecosystem built for them, and accumulated data and workflows will become advantages that competitors find difficult to replicate in the short term.
As in the field of images, understanding and serving users’ visual expression is also at the heart of video. Meitu’s approach has been to capitalize on the industry trend of orchestrating models into workflows, quickly identify high-value vertical scenarios, go deep into those scenarios, and continuously optimize through user feedback to deliver better results.
Continuing with this approach, Meitu launched a native AI video creation platform, MVLAND, in June this year.
In terms of product features and workflows, the platform almost entirely reflects the latest AI video approach: through a music analysis Agent and a creative canvas, it precisely transforms the rhythm and emotion of music into visual imagery. At the same time, MVLAND continues Meitu’s distinctive vertical-market strategy by focusing on the music video creation workflow and delivering high-quality music videos for musicians, content creators, and others.
MVLAND not only achieved PMF (product-market fit) in a short period of time, but also demonstrated Meitu’s ability to build highly monetizable products. According to Meitu, MVLAND doubled its ARR within three months of launch. At the same time, its average monthly ARPPU (average revenue per paying user) reached approximately RMB 220—13 times the combined ARPPU of Meitu’s various previous products.
This shows that when a product can truly deliver high-quality results, help users solve important problems, and even support them in achieving commercial conversion, users become far less sensitive to price.
From Wink’s debut in 2022 to MVLAND’s launch in 2026, the market’s perception of Meitu has gradually shifted from “a company that only retouches photos” to “Meitu can make videos, too.” As video becomes one of the AI industry’s most definitive growth areas, Meitu’s deep understanding of visual expression is transforming from a soft capability into a hard barrier to entry.
A more lucrative business than images has already proven viable, and the possibilities embodied by video may only now be beginning to unfold. It may be time to reassess the market’s valuation logic for Meitu.
This article was provided by Meitu and reprinted by QbitAI with authorization. The views expressed belong solely to the original author.