Meitu’s AI business is moving towards delivery.

On June 17, the 2026 Meitu Imaging Festival was held in Xiamen, releasing four new products: Picchi, Artflo, MVLAND, and MeituHub, while also upgrading Zcool, Meitu Design Studio, Kaipai, and RoboNeo.

Looking at Meitu's AI upgrades over the past two years, this change focuses more on outcome delivery.

In the past, Meitu offered functions around photo editing, shooting, and design. This year, product features are clearly shifting toward deliverable results: portrait retouching, dubbed videos, marketing materials, music visuals, and AI short dramas are all packaged into deliverable workflows.

CEO Wu Xinhong said in a group interview that user needs are changing in the AI era. As more AI applications become available, many users don’t actually want to spend time learning them. Corresponding to this year's products, users may not want to learn prompts, parameters, or complex workflows; most paid behaviors focus on directly usable images, videos, and commercial materials.

From Agent to Agent Teams, From “Delivering Functions” to “Delivering Outcomes”

Picchi focuses on deep portrait retouching, MVLAND targets musicians and labels, Kaipai is for dubbed and marketing videos, Meitu Design Studio is now building AI design teams, and RoboNeo is exploring AI short dramas. These products belong to different scenarios but point toward the same commercial direction: fewer function buttons, more direct outcomes delivered to users.

Goldman Sachs’ recent report on Meitu summarizes its transition from a “beautification tool” to a “professional content engine.” Generative AI is driving Meitu’s expansion from consumer entertainment to enterprise productivity tools, projecting a compound annual revenue growth rate of 29% from 2025-2030, with productivity tool revenue rising from 12% in 2025 to 44% in 2030.

At the launch event, more direct signs of commercialization came from two sets of data shared by CFO Yan Jinliang.

Some Meitu Design Studio users’ AI computing consumption has reached over ten thousand yuan per month, and MVLAND recently saw a monthly ARPU of three to four hundred yuan—about 20 times that of Meitu Xiuxiu.

For Meitu, which has long relied on C-end imaging tools as its mainstay, these figures are more informative than product introductions, revealing that AI productivity products are creating new payment tiers.

Meitu develops each scenario into a concrete product delivered to users: Kaipai serves dubbed and marketing videos, Picchi serves deep retouching users, MVLAND serves music visuals, RoboNeo serves short drama creators.

Regarding this, Wu Xinhong explained, “Each of our products has its own positioning; we want to perfect each scenario.” He also mentioned that Kaipai intends to "go all out" in the dubbed video field and ensure it becomes a leader globally.

This product segmentation is related to the commercialization logic of AI applications.

General models can provide underlying generative capabilities, but the paid value of imaging products often appears in specific scenarios. Users pay for model invocation, aesthetic standards, templates, materials, workflows, industry insights, and final results.

J.P. Morgan's China AI application research report states that in some enterprise scenarios, access to foundational models might become more interchangeable, and customer value depends more on task completion, workflow integration, proprietary data, and deployment quality.

For Meitu, model capabilities are just entry requirements; whether a product can become part of users’ workflows determines retention and payment ceiling.

Technically, Meitu’s “Agent Teams” solution allows multiple single agents to collaborate and verify in multiple phases, enabling complex workflows and better commercial outcome delivery.

Meitu’s vertical products also derive from needs in mature products. Kaipai originated as a teleprompter in the Meitu Camera, and Meitu Design Studio came from poster design features in Meitu Xiuxiu.

Meitu CPO Chen Jianyi said in a group interview that the commonality of these two products is clear demand and a deeper understanding of users. Many on the Kaipai team are content creators themselves, and Meitu Design Studio identifies pain points through offline exhibitions and user research.

Models Enter the Business Closed Loop

In discussions about "models eating applications," Yan Jinliang responded during the group interview.

He believes this judgment doesn't hold true in the visual domain because subjective aesthetic preferences require products and services for end-stage calibration. A model can generate an image, but whether users find it natural, premium, publishable, or able to drive sales still depends on aesthetic standards, product refinement, and real feedback.

Meitu has made organizational adjustments for this. Design Center head Xu Jun introduced that the company has a design center with over 200 staff. Designers first create samples for AI, teach aesthetics and short drama knowledge to AI, then feed user feedback back into AI training. Designers' deliverables have expanded from images and videos to models, creative calls, and workflows.

Model selection now follows business validation. On-site data showed that from January to May 2026, an average of 96.3% of generative AI calls in Meitu imaging products came from Meitu’s own Qixiang large model.

MT Lab head Liu Luoqi said in an interview that self-developed large models are more closely integrated with the company’s business products and serve highly customized scenarios; relatively fixed scenarios may use third-party large models.

Chen Jianyi provided a more practical formula: "Model R&D depth x the business value of the corresponding product." If a scenario hasn't yet proven its payment ability, external models can be used to experiment; when commercial value is established, then invest in self-developed model capacity. This order of investment is closer to the operating logic of application companies than simply stressing self-development.

AI’s impact on Meitu’s business will ultimately be reflected in its business model. Previously, Meitu mainly charged for tool functions like photo editing, filters, templates, and video enhancement. Now, productivity scenarios are bringing subscriptions, AI computing value consumption, real-person creation services, and result delivery around industry needs.

UBS’s recent report noted that AI will increase the interaction frequency between software vendors and customers and lower delivery costs, supporting accelerated software demand and improved profit margins over the next 3-5 years. Traditional software companies need to choose between AI-enhancing existing products and rebuilding AI-native products.

Within this context, the 100 million yuan product challenge also has business significance. Meitu has 280 million monthly active users, a mature imaging product matrix, and Zcool’s 18 million designer resources. However, imaging segments are fragmented, and new AI-native products will keep emerging.

Incorporating external innovation into the product funnel is, in a way, supplementing its imaging ecosystem with more vertical scenarios.

From this imaging festival, Meitu’s AI transformation has already moved from “adding AI to products” toward “using AI to reconstruct products, organizations, and charging methods.” Next, the market will focus more on whether these applications can consistently deliver payable imaging outcomes. If products like Kaipai, Meitu Design Studio, and MVLAND continue to prove the replicability of high ARPU, Meitu’s valuation anchor will shift from imaging tools to imaging productivity platforms.

This article comes from WeChat Account “Hard AI”. To follow more AI frontier news, visit here.

Risk Warning and DisclaimerThe market involves risks, and investment should be cautious. This article does not constitute personal investment advice, nor does it consider individual users’ special investment goals, financial circumstances, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their specific situation. Investing accordingly is at your own risk.