Reported by |
Edited by | NUPIAO
To stay ahead of being swallowed up by general-purpose large models, Meitu is strategically rolling out a whole new lineup of imaging tools.
During its annual Imaging Festival, Meitu unveiled four fresh apps—Picchi, Artflo, MVLAND, and MeituHub—alongside eight brand-new AI features. We’re talking designer communities, portrait retouching, commercial graphics, talking-head videos, concept art, music visualization, AI short dramas, and full-scale AI image pipelines. Every single one slots into a different corner of the content creation workflow.
With AI shaking up the industry, Meitu has quietly shifted its compass over the past couple of years, firmly locking itself into the “AI application” lane. Dropping so many AI tools at once might look like a scattergun bet, but there’s actually a method to the madness.
When asked about this move, Meitu founder Wu Xinhong explained it straight: they don’t want to dump everything into one bloated app because each tool needs to own its own niche and nail it. He’ll admit that we won’t know which of these four new apps will really take off until later, but think of it as a funnel. The ones that survive user testing will rise to the top.
So how do they widen that funnel? By mining internal feedback and scouting outside projects. Chief Product Officer Chen Jianyi shared that Meitu taps into its mature apps, AI innovation labs, and product hackathons to spark new ideas. Take Picchi, for example—it basically spin-offs power users from Meitu Xiuxiu into a dedicated space. Honestly, tons of their hits came straight from user requests; even their early “teleprompter” feature in the BeautyCam app became a top-three revenue driver before evolving into the KaiPai app.
On the external front, Meitu is throwing 100 million RMB into an incubation fund at this year’s festival. They’re challenging AI devs across the board with a product contest, handing out up to 5 million RMB in seed funding for the finalists.
That said, when picking which lanes to run, Meitu is actively hunting for blue ocean pockets instead of fighting over crowded spaces.
Wu pointed out that beyond tracking their own internal pipeline, they’re keeping a sharp eye on “unconventional” bets. MVLAND, for instance, is built specifically for musicians and indie labels who need fast, high-quality MV production.
Meitu also rolled out an AI short-drama maker at the festival. Sure, the domestic market’s already boiling over, but they’re planning to crack the overseas door first and fold this hot sector into their funnel. As Wu noted, AI short dramas demand serious filmmaking chops, and the global supply chain for this format is still wide open. That leaves plenty of room to build a world-class AI tool for international creators. This lines up perfectly with reports that Meitu’s been aggressively hunting for niche opportunities abroad.
You can’t ignore the underlying anxiety driving these moves. General-purpose consumer AI apps are grabbing massive audiences, and everyday photo editing tasks are slowly drifting away from vertical apps. At the same time, Meitu still leans heavily on image and design subscriptions, but monthly active user growth has clearly stalled—reaching just 276 million globally in 2025, a sluggish 3.8% year-over-year jump.
Wu doesn’t dismiss the impact of the tech giants, acknowledging that their base model capabilities definitely dictate how downstream apps perform. But he’s quick to point out that raw foundation models still struggle to deliver polished results in highly specific verticals. The real “last mile” of imaging requires fine-tuning parameters, crafting visual assets, managing content, and nurturing creator communities. Big traffic doesn’t automatically win here. In fact, most of the fastest-growing players in this space globally are still nimble startups—and that’s exactly why we’ve doubled down on acting as a startup incubator ourselves.
This year’s launch isn’t just about tweaking old AI features—it’s heavily pushing AI Agents and focusing squarely on “deliverables.” Underneath all that hype, the biggest shift is actually financial: Meitu is transitioning from a traditional subscription model straight into pay-per-AI-compute-token pricing.
The numbers back this up. According to earlier disclosures, AI compute token spending for their imaging and design suites jumped 59% by March 2026 compared to December 2025. Live data from the festival shows even crazier spikes: KaiPai’s token consumption skyrocketed over 7x by May, and Meitu Design Studio’s usage exploded 8x over just eight months.
Of course, scaling this fast comes with a steep price tag. Higher compute loads and marketing push mean rising costs—a clear reality check for Meitu’s aggressive expansion. Their 2025 earnings report actually flagged gross margin compression, largely driven by those climbing cloud and GPU expenses.
CFO Yan Jinliang spilled some numbers on the compute side: between January and May this year, a staggering 96.3% of generative AI calls across Meitu’s imaging apps ran on their own Qixiang LLM. Yeah, newer or beta-stage tools still burn through more third-party model credits, but as those products mature, Meitu’s custom vertical models and design center infrastructure will steadily cover their needs. Frankly, leaning hard on external models was never part of the long-term game plan.
In short, whether they’re chasing revenue growth or squeezing down compute costs, it all boils down to one thing: can Meitu successfully filter out winners through that product funnel? For a company that’s spent years coasting on proven, mature apps, this is honestly a reset button moment—they’re sprinting back to the starting line, trying to find their next breakthrough.
