For nearly two decades, Meitu has served as the definitive mirror for China’s digital vanity. From its origins as a PC photo-editor to its dominance as a mobile-first 'beauty' ecosystem, the company has survived every major platform shift in the Chinese internet landscape. However, the rise of Large Language Models (LLMs) presents a different kind of challenge: the possible obsolescence of the application itself. As AI agents begin to execute complex creative tasks through simple voice commands, the traditional 'toolbox' interface of icons and sliders is increasingly viewed as a relic of a dying era.
In a candid reflection on this shift, Meitu Chairman Wu Xinhong describes a future where the company must 'tear down its walls' to survive. The strategy marks a radical departure from the walled-garden approach of the mobile internet era. By integrating its core imaging capabilities into open frameworks like OpenClaw, Meitu is transforming its proprietary aesthetic algorithms into a public utility. This move acknowledges a harsh new reality: in the age of AI, a product can no longer exist as an island; it must become a 'skill' that other models and agents can call upon.
This transition from a destination app to an infrastructure provider is reflected in Meitu’s evolving balance sheet. Revenue is increasingly tied to 'Tokens'—the digital currency of AI processing—rather than traditional subscriptions alone. When third-party developers or rival platforms use Meitu’s beauty filters via API, the company collects a toll. This pivot highlights a strategic retreat from the 'model wars' that have drained the coffers of tech giants. Having briefly experimented with building its own foundation models, Meitu pivoted to a 'model container' strategy, utilizing a mix of open-source and third-party APIs to keep costs manageable while focusing on vertical expertise.
Internally, the company has restructured itself to mirror a venture capital firm rather than a traditional software house. Meitu now seeds dozens of internal 'AI studios' with small amounts of capital, encouraging a culture of high-velocity failure. If a product like the recently shuttered MOKI—an AI short-film tool—is outpaced by giants like ByteDance or Kuaishou, Meitu cuts its losses immediately. This decentralized approach is a recognition that the 'product-market fit' of the AI era is too volatile for top-down management to predict.
Despite the encroaching power of generalized agents, Wu Xinhong maintains that vertical specialization remains a moat. While a general-purpose LLM can retouch a photo, it often lacks the nuanced 'aesthetic judgment'—the precise skin tones and lighting preferences—that Meitu has refined over eighteen years. The company is betting that even in a world of super-intelligent agents, there will always be a need for specialized 'expert agents' that can provide stable, high-quality results in specific domains like professional imaging and fashion.