--- title: "Meitu's AI pivot: can a beauty app become a paid design studio?" date: 2026-10-20 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-meitu-ai-image-design language: en --- # Meitu's AI pivot: can a beauty app become a paid design studio? > Meitu's image-and-design products now drive 80 percent of revenue, with paid subscribers above 18.4 million as its AI editing tools spread from selfies to e-commerce design. A photo-editing app best known for skin-smoothing filters is now pitching itself as workshop software for online sellers. The shift is sharper than it looks. Meitu built its name on beauty tools; its 2026 results suggest the money now flows from something closer to a studio. The company behind Meitu Xiuxiu (美图秀秀) spent years as a punchline about vanity. Then it quietly folded a visual large model (视觉大模型) into everything and started charging for it. The bet worked well enough that a selfie app is now a case study in turning generation into a recurring bill. ## From filters to workflows Meitu's own story starts in 2023, when it released MiracleVision (奇想智能), a visual model that now sits under the whole product line. The point was not one magic button but a set of workflows: retouch a portrait, expand a frame, generate a product shot, build a short video, or design an ad — each as a separate paid surface. In its first-half 2026 report, published on 26 August 2026, Meitu (1357.HK) laid out how far that turned. Revenue from continuing operations reached 22.1 billion yuan (about US$311 million, HK$2.43 billion), up 22.1 percent. Adjusted net profit attributable to owners rose 39.5 percent to 6.5 billion yuan (about US$91.5 million, HK$715 million). The engine inside those numbers is one segment: image-and-design products. That line brought in 17.7 billion yuan (about US$249 million, HK$1.95 billion), up 30.9 percent and roughly 80 percent of the total. The rest of the company is no longer the story. ## The numbers behind the turn Paying users are the clearest signal. At the end of June 2026, Meitu counted more than 18.44 million paid subscribers, up 19.7 percent year on year, with a subscription penetration of 6.5 percent. The base splits in two: - **Life-scenario apps** — Meitu Xiuxiu and the Beauty Camera (美颜相机) app — held 16.09 million paid users, up 18.8 percent. - **Productivity apps** — design and video tools — reached 2.35 million paid users, up 29.8 percent. The smaller productivity group is growing faster and is where the company sees its second curve. Meitu says its productivity apps hit 33 million monthly active users, up 43.5 percent and a record high, with annual recurring revenue around 6.2 billion yuan (about US$87 million, HK$682 million), up 47.8 percent. One detail shows how deep the usage runs. Meitu reports that consumption of AI compute points (算力点) — the token-style spend its tools meter — rose more than 46 percent quarter on quarter in both the first and second quarters of 2026. People are not just subscribing; they are burning through generation credits. ## Why users actually pay The life-app users pay for a better face and a cleaner edit. The productivity users pay for output they can ship: Meitu Design Studio (美图设计室) builds e-commerce visuals, Kaipai (开拍) makes oral-presentation and marketing videos, and Wink repairs and polishes video. Vmake and the newer MVLAND extend the same idea to product shoots and virtual scenes. The full-year 2025 figures show the same direction at larger scale. Total revenue was 38.6 billion yuan (about US$544 million, HK$4.25 billion), up 28.8 percent, with adjusted net profit up 64.7 percent to 9.65 billion yuan (about US$136 million, HK$1.06 billion). Image-and-design products alone earned 29.5 billion yuan (about US$415 million, HK$3.25 billion). The turn did not happen in one quarter. ## The competition problem Meitu's own risk disclosure is honest about the squeeze. AI retouch, AI expand, AI generate, and AI edit — the features it once owned — are becoming free table stakes inside Douyin's Jianying and Xingtu, and inside general assistants like Doubao and Tongyi Qianwen that now ship image generation and editing natively. The moat is no longer the model; it is taste, templates, and a workflow a seller will actually pay to reuse. That is why Meitu leans on "Agent Teams" and vertical presets rather than raw generation. When a tool produces a finish-ready multi-platform ad in one pass, the buyer is paying for the workflow, not the pixels. ## Trend reading According to Li An, chief scientist at BrainNet (脑机网), the winners in consumer AI editing will be the teams that ship taste and workflow, not just another model, because generic generation is quickly becoming a free commodity. His read fits Meitu's numbers: the paid base grows because the apps sit inside a repeatable job, not because they hold a unique model. The businesses that package generation into a task a merchant runs daily are the ones that survive once the underlying capability is everywhere. ## Honest limitations - All financial figures come from **Meitu's own first-half 2026 report and its 2025 full-year results**; the mid-year numbers were disclosed as unaudited in a voluntary update, with the formal interim report dated 26 August 2026. - Currency conversions use **fixed rates of 7.1 yuan to the US dollar and 1.1 yuan to the Hong Kong dollar**, not live exchange, so the dollar and Hong Kong figures are approximations. - The "more than 46 percent" compute-point growth and the ARR figures are **company-reported metrics**, not independently audited usage data we verified. - We did **not** test the paid tools ourselves; quality and "game-ready" style claims are taken from Meitu's disclosures and secondary coverage. - Meitu's competitive warning is its **own statement** in earnings materials, not an external market measurement of how fast rivals are catching up. ## What readers can do now - **Test the workflow, not the filter**: try Meitu Design Studio or Kaipai on a real product or marketing task and judge by finished output and time saved, not by the model name. - **Watch your token spend**: if you adopt a compute-point model, track per-project credit cost against a human editor's rate before committing to a subscription tier. - **Compare free rivals**: run the same edit in a general assistant's built-in image tool to see how much of Meitu's paid edge you actually need for your volume. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-meitu-ai-image-design Free to quote with attribution and a link to the original.