Good afternoon, everyone. Welcome to Meitu's 2026 Interim Results Presentation. Simultaneous English interpreting will be provided throughout the meeting. The management present today include Founder of the Company, Chairman, and CEO, Mr. Xinhong Wu. CFO and Company Secretary, Mr. Gary Ngan, a nd CPO and President of the Imaging Products, Mr. Jianyi Chen. A warm reminder for you guys. Today's speech may include some predictive statements. Many risky factors we cannot control may lead to discrepancies between actual results and our statements. For online participants, please click rename button with your real name and affiliation name. After the Q&A session, please state your name and affiliation before raising your question. For online participants, you can question via audio or text. Now, let's welcome the management to share about the updates of the company. Thank you.
Thank you. Good afternoon. Thank you for joining us today for our interim results presentation. In the first half of 2026, our continued focus on vertical use cases translated into tangible results, driving growth in our core businesses and further improving profitability. To date, the goals of 2026 Meitu Multimedia Festival have been delivered on schedule. Applications for productivity, such as Designkit and Kaipai, continue to unlock commercial value. New products including Picchi and MVLAND, have also brought growth opportunities. We are very optimistic about the opportunities at AI application layer. For Meitu, this is an important window. Based on this, in the first half of the year, we continued to upgrade our existing products, including accelerating deployment of agent teams across our applications for productivity, so that agents move beyond individual tasks and collaborate across workflows. We are already seeing the early results.
In both the first and second quarters of this year, total AI credits consumed by users increased by more than 46% sequentially. This indicates continued improvement in both user engagement and willingness to pay. At the same time, user needs and product formats continue to evolve. To capture this window, we need to accelerate iteration of existing products, and also identify and validate new opportunities more efficiently. Last year, we started our internal AI innovation studio. Two months ago, we also launched the Meitu Hatch Catch to find high-potential AI imaging products. To date, we have received 109 submissions, and we will select those suitable for strategic investment or in-depth collaboration. In overseas markets, we encourage our product teams to get closer to local user and use cases.
In markets such as Brazil and Mexico, we're deepening our understanding of local content creation habits, the relationship between content and consumer acquisition, and unmet user needs. These new product explorations and market studies showed us a clear trend, that is, more and more users are using social media to build their personal IP and branding in order to acquire customers and monetize. What these users need is no longer one-time creation. They need to continuously create content and operate their IP, expand their social influence, and build reusable assets. This showed us that beyond content creation, there may be opportunities to, based on user needs, extend further into what comes after creation. I would also like to share some of our thinking on this. In summary, we hope to, around content, connect creation, distribution, and asset accumulation. First, AI content creation.
This is the starting point of content and the area we know best. Our products have already covered multiple formats, including marketing videos, MV, talking videos, and AI short dramas. We hope that while focusing on creation, where we are strongest, we can expand our user base and content scale. Second, AI content distribution. This is where content reaches audiences and creates commercial value. Once the creative work is completed, the next question is how to reach more audience and unlock its commercial potential. On distribution channels, we have our own ones, including Meitu app Portfolio and ZCOOL, while connecting to external channels such as social media, video platforms, music streaming platforms, and offline stores. While we accumulated many users, we are also finding high potential content at an earlier stage. Third, the AI asset library, the foundation for content accumulation.
We want content to remain usable after creation distribution, and thus generate long-term value. They can take many forms, including reusable materials, long-term IP and virtual characters, as well as copyrights of music, scripts, and visuals. This is not a new initiative. These capabilities have already emerged in our product portfolio in different forms. Take Kaipai as an example. A talking video creator uses Kaipai to produce a video. That is creation. Once completed, they want more people to see it. That is distribution. Going forward, we can support the journey from creation to promotion within the product.
Later, the creator's persona, voice, and script templates can now be reusable, and the next video comes faster and more consistent. MVLAND, which we launched this year's Meitu Multimedia Festival. A music studio uses MVLAND for creation. The content then goes into streaming platforms and social media for distribution.
The copyrights and virtual identities accumulated through each creation become assets. These assets can be licensed and reused while virtual artists can build their audiences and commercial value, thus creating a positive cycle. We will, based on this, continuously expand and improve MVLAND. In the future, we will continue to connect the capabilities more systematically, starting with where we have already saw clear demand and positive signals. Building on our strengths in AI creation applications, Meitu will extend into AI content distribution and asset libraries, enabling content to be created, circulated, and accumulated within our ecosystem, building a value chain for the AI era. Sure we will do it step by step. Next, we will continue on our existing products and explore vertical use cases, while through more efficient innovation mechanisms, we will try to identify and rapidly validate new growth opportunities.
Whether through internal development or strategic investment, we will consider both business synergies and return rates, and focus resources on areas with the strongest value and potential. Going forward, we will continue to capture the window for AI applications while balancing growth quality and operating efficiency, driving sustainable growth, and creating long-term value for shareholders. Next, I will hand over to Xiaobai, our Chief Product Officer, share some of our latest product thinking and progress with specific cases.
Thank you, Xinhong. Next, with a few specific products, I will share some of our recent new insights. First, two of our fastest-growing products after the festival, Picchi and MVLAND. As of now, both products' ARR has passed $0.5 million . They are among the best. Picchi has grown fast since its launch. This year, it has a chance to reach $1 million in ARR. More worth sharing than the numbers is its users. Today, most of Picchi's users are cosers, cosplayers, namely. This group retouches a huge amount, dozens or even hundreds of photos from one shot, different from C-end users. They care most is not exactly whether it looks like me. They are sensitive to that point. Cosplay by nature is playing a character, so they focus more whether the final outcomes has quality and could be shareable on social media.
Because for many professional cosers, image quality directly affects follower growth, ad deals, and commercial shoots. Many cosers treat this as a job, not only a hobby. This gives us a very important hint. When personal image itself becomes a means or material of production, users' demand for imaging products and willingness to pay will both change. Meitu app and BeautyCam more serve later scenarios. The core value is still make me look better. Picchi is not for mass end users, but for those who monetize their personal image like streamers, influencers, models, cosers, and so on. For them, a good-looking image is not just about consumption, but production. So their willingness to pay is naturally higher than C-end users. These are our observations on Picchi. Next, MVLAND. It now has the second highest ARR among the company's new products after Picchi.
It might be very niche, but now it seems to be different. It could be creative work, and it can also be marketing content. MV could be work and also advertising or marketing materials. For musicians, traditional MVs cost a lot and take long. The result, the vast majority of songs have no MV and can only get buried on traditional music platforms. Today, the key place for music discovery has shifted to social media. MV is exactly the ticket for a song to be seen. MVLAND lowers the production cost, which is like giving more songs a chance to be distributed. This is a form we have tried for the past several months. Another type is the marketing MV. We have a saying inside, "Everything can become an MV." When marketing, it could be spoken or it could be sung. Kaipai, focusing on talking videos.
We can speak it out, we can talk about it. For normal sellers, they can also utilize AI to produce a short MV and communicate it on the social media for a milk tea shop or making the advertising materials for new clothes. They can use this interesting way to promote. More in the past, it could be a static poster, but I believe they could use this in the future. We hope to turn MV from a professional content format in the music industry into a marketing form that everyone can use. This also links to what Xinhong just said on distribution. Creation and distribution are both essential. Without good content, distribution has no ammunition. Without effective distribution, good works are also hard to be seen. So what MVLAND wants to do is connect creation and distribution and build relevant asset accumulation.
For example, the virtual persona, so that every song can have the ability to be seen, spread, and marketed. Third, Kaipai. One insight we formed recently, the creator density. After the festival, we went to São Paulo for research. At first, we watched TikTok locally. The likes are central, but the granularity may not be very fine. We wondered, are there less creators in Brazil? But after our research back in China, the truth is totally the opposite. There are so many creators in Brazil, but what's truly scarce is professional creators who can make money from content. So now we distinguish two metrics. One is lay creator density, one is professional creator density. The lay creator density, you can share your life, but they don't monetize from their content. For professional creator density, they need to rely their living on the content they make.
If there are many creators in the market, but just recording their life, they have a chance using Meitu's app to become professional. So when lay creators are many, while professional creators are less, this is our chance. This is why Kaipai's growth in China is still good. For healthcare, insurance, real estate, and food industry, they know how to run their own business, but they may not understand how to select their topic or advertise them. They don't know how to run a social media account. We give them a chance. They may not use professional tools, but they could be quasi professional content. The leading indicator is not population, but creator structure. Places with large base of labor creators and many people wanting to go professional and monetize content are our opportunity points. Finally, Designkit.
One of the major changes is the shift of user habits in the past two or three years. The e-commerce content is moving from shelves to social media. In the past, consumers saw a product on social media, then went to an e-commerce platform to view the detail page. Their decision focused on shelves e-commercers like Amazon, but now it is totally reverse. They see a product on social media, and they go to the live streaming. They just decided to buy. Social media can provide a whole chain for purchasing on. More and more purchasing decisions happen on content flow in social media. This is directly impacting our Designkit product. In the past, the largest group serves is cross-border e-commerce merchants. Their operating model is extending. No matter it is about the agent or the cutout function, we centered around the shelf-focused e-commercers.
The main picture or the detail page. But now they need more pictures, texts to form their account metrics. In the past, e-commercer may need to spend 80% on traditional image, 20% on social media. Now the proportion is reverse. This is a new designing point for us. Designkit will take two steps. First, make the capability for e-commerce material solid, and on that basis, build up accumulation of product, brand, and model assets. This is what we have achieved for the first phase. Second, we extend to social commerce and around two types of content, product and service. As the starting point, we help merchants bring business into social content feed. Put these products back into the framework Xinhong just described. In fact, creation, distribution, and accumulation, these three stages have all started to appear in our products.
On the creation side, the Meitu app and BeautyCam are the most popular creation points, further solidifying our user scale. Picchi, MVLAND, Kaipai, and RoboNeo extend creation capability into personal images, persona MVs, talking videos, and short dramas. These daily update content forms the growth driver for productivity scenarios in the coming years. On the distribution side, Kaipai extends from creation to promotion. MVLAND connects MV creation and distribution. The Meitu app also keeps helping brands and IP expand exposure. For filters, ZCOOL lets design content be seen and traded in its creator community. On AI asset accumulation, every creation and circulation is accumulating.
For example, Picchi users' image styles, MVLAND's music copyrights and virtual singers, DesignKit's brand and product assets, ZCOOL's years of design works, all are entering the asset library, being called again and again, and feeding back into the next creation and distribution, building Meitu's complete content value chain in the AI era. This also enhances our foundation from productivity tools to distribution, to communication and make it deeper. Now let's welcome Gary to share the operating and financial performance of the first half.
Thank you, Xiaobai. Let me start with a few key metrics. As of June 2026, our overall MAU was about 282 million. Applications for productivity reached 33 million, record high. Applications for leisure remained stable. Paying subscribers exceeded 18.44 million, among which applications for productivity reached 2.35 million, up by nearly 30%. Applications for leisure reached 16.09 million, up by more than 18%. Our overall user base remains stable, while for application productivity, both the user scale and paying subscriber base continue to grow. A few ARPPU highlights. The ARPPU for productivity application is currently 150% than of the leisure application, but we believe there is still more headroom in the future. The ARPPU gap reflects the limitation of the traditional subscription model. That is a cap on user spending. Even when users are willing to pay more, the product lack mechanism to capture that.
In the first half of 2026, we integrated the agent capabilities across our products, lifting the limit. As users approach more complex tasks at higher rate, they will consume more AI credits. This is already visible among high-value users of single products. For example, Designkit already has several thousands high ARPPU users whose annualized spending is approximately 10 x of the total, while the number of such users has increased by 6x compared with December last year. Another example is MVLAND. Within only a few months, its average monthly ARPPU has already reached approximately CNY 220, comparable with pricing levels for professional creative applications. It demonstrates that we are capable of building products that monetize at a much higher level. Now let me return to financial results. First, let me clarify the basis of our comparison.
As the cosmetic supply chain management services business under the solutions for beauty industry was classified as a a discontinued operation in our 2025 annual report, the following year-on-year comparison are presented on a continuing operation basis. In H1 2026, the group recorded total revenue of CNY 2.21 billion, an increase of 22.1%. Within this, photo, video, and design products grew by 30.9% to CNY 1.8 billion.
By product category, revenue from applications for productivity accounted for more than 80% of the photo, video, and design products. These growth rates exclude the stock photo business that we exited at the end of 2025, which contributed approximately CNY 27 million of revenue in the first half of 2025. Revenue from applications for leisure accounted for approximately 82% and increased by more than 30%. By market, revenue from markets outside of mainland China increased by 41.7%, accounted for 48% of the total revenue.
For our advertising business, revenue in the first half was CNY 410 million, down 4.4%, primarily due to lower revenue from brand advertising and other traditional advertising formats. As we have mentioned in previous presentations, we remain cautiously optimistic about the advertising business. Going forward, we will continue exploring opportunities to combine AI capabilities with brand marketing. Finally, revenue from others was approximately CNY 29.1 million, increasing by 3.9% year-on-year. For cost sales, for the first half, it was CNY 630 million, up by 41%. Our largest cost item, revenue sharing fee to payment channels, was CNY 380 million. This increased alongside subscription and in-app purchase, which remained broadly stable as a percentage of photo, video, and design products revenue. Second, computing and cloud costs accounted to CNY 130 million, over half of which was computing costs directly linked to usage.
In addition, supported by our self-developed models and our capabilities to fine-tune open source models, we kept third-party API costs at a mid-single digit percentage. During the first half, more than 96% of the generated outputs were powered by our self-developed model layer. Turning to gross profit. For the first half, it was CNY 1.58 billion, up by nearly 16%, while gross margin was 71.5% compared with 75.3% in the same period last year. This is mainly attributable to changes in revenue mix. First, within photo, video, and design products, AI-native applications productivity grew faster, and they now have higher API cost ratio. Mature applications maintained stable gross margin levels. For these early-stage AI-native products, our current priorities are product validation and user growth. As user scale expands and patterns become more mature, we will improve their growth margin profiles.
Second, advertising which carries a higher gross margin, accounted for a lower proportion of the total revenue, which also had some impact on overall margin. For expenses, selling and marketing expenses were CNY 330 million, up by 12.7%, representing 15% of the combined revenue. By origin, spending in mainland China was primarily for driving growth in applications for productivity, while in international markets, it was mainly for applications for leisure. For products with validated demand and clear growth opportunities, we will continue to invest and monitor customer acquisition efficiency.
R&D expenses were CNY 473 million, up by 4.8%. The growth rate was affected by comparison base in first half 2025. Which included certain general purpose foundational model training expenses. Including this, the total expense would be increased approximately 11.1%. Going forward, we will continue to concentrate resources on R&D talent, vertical model training, and application-level optimization.
Administrative expenses were CNY 223 million, largely in line with last year. On profitability, as revenue continues to grow, the gross profit once again grew faster than operating expenses in the first half, further improving operating leverage and overall profitability. Adjusted net profit attributable to owners of the company was CNY 6.5 billion. For better comparability with last year, adjusted net profit includes results from discontinued operations. The IFRS net profit attributable to owners of the company was CNY 620 million, representing an increase of 40%. I would also like to share our approach to investment in new products and resource allocation. You see that we have launched quite some new products recently. Some, such as Picchi and MVLAND, are growing rapidly, while some have been discontinued. This reflects the speed of evolving AI industry, and therefore the speed of validation.
For early-stage opportunities, we try to use small teams and limited initial investment to bring products to market quickly and validate user demand and business models. Once we see clear signals in user growth and monetization, we will scale up R&D, marketing, and other resources. Through this approach, we aim to maintain a rapid pace of iteration while controlling the cost of exploration and improving resource efficiency. Going forward, we will continue innovation and globalization, capturing long-term opportunities while maintaining good resource allocation and investment returns, and therefore healthy financial performance.
Finally, shareholder return. First, in line with our past practices, the board has approved an interim dividend of HKD 0.067 per share, representing 40% of adjusted net profit attributable to owners of the company. Second, regarding the HKD 300 million share repurchase program announced in March this year, we have repurchased about HKD 230 million, with HKD 70 million remaining.
Going forward, we will assess and execute remaining repurchases as appropriate, taking into account the market conditions and other requirements. I would also like to clarify our principles regarding results to disclosure. Recently, we have noticed that some investors interpreted whether the company issues a positive profit alert or profit warning as an indication of management view on whether the results are good. Actually, there is no link. Under the requirements related to the inside information based on Part XIVA of Securities and Futures Ordinance and Rule 13.09 of the listing rules, whether the company is required to issue a positive profit alert or profit warning in advance does not depend on whether the results themselves are positive, but whether they constitute inside information. In other words, whether the annual interim results represent a material deviation from prevailing market expectations.
Since we began publishing quarterly disclosure this year, the market has been able to learn more about our latest operating performance. The likelihood of actual results deviating also decreased, and accordingly, there may be fewer circumstances requiring a positive profit alert or profit warning. Therefore, having an announcement or not should not be interpreted as an indication of the results. We will continue to fulfill our disclosure obligations in accordance with the rules. That concludes our presentation today. We will now move on to the Q&A session. Thank you.
Thanks, Gary, for his sharing. Now we move into the Q&A session. If you are on the floor, please raise your hand. Our staff will give you the microphone. If you log in via mobile app or personal computer, please click Raise Hand button. After you are unmuted, you can speak your questions. You can also text your question in the chatting box. We will read it aloud. For investors in Zoom, you can also click the button or text your question. For those who dial in, please press asterisk plus one. First question on the floor.
I am [Analyst] from Swiss Bank. Congratulations on company's performance. I have two questions centering on one topic. In the first half, we developed so many new products, Picchi and MVLAND, especially MVLAND. We can see it breaks the boundaries or extends our functions.
I would like to ask how we discovered so quickly these demands. I saw the ARPPU value and the ARR value are both impressive. Second question about the verifying mechanism of new product. What KPI decides the continuation of a project? For the internal talent accumulation, do we have boundaries in the future, or we are just exploring? Thank you.
Management answers. First, I talk about the discovery of demands of MVLAND and Picchi. For MVLAND, 12 years ago, Meipai, we already discovered it, but the main difference is that Meipai based on users' real experience. After its launch in May 2014, the download reached 5 or 6 million. So it proved one thing, users like to make MV gathering their life materials. You can regard MVLAND as a Meipai in AI era.
For the initiating phase, it targeted professional musicians to solve their problems because we adopted different approaches. We want to first serve the musicians well, and we want to move downward. For example, the MV could satisfy the needs and aesthetic values of the professional musicians. We believe it could serve more audiences or population. It is not a brand-new scenario. This is what we want to define the content format. We can use the form of music video to do more, even if it might seem it is a bit niche nowadays or in the past. For example, leveraging the advantage of AI models. Xinhong asks how it digs these demands? From our existing products, we discover new trends or demands. For example, Picchi. We started from the Meitu app and BeautyCam and found young users or KOL, they are dealing on portrait photos.
In this process, how AI could solve their pain points, how to remember their preferences or different traits to do many pictures at one blow. Or we could make it more sensitive to social media. Besides Picchi and MVLAND, for example, Kaipai and other, we used the prompter function to discover users' demands. We will continuously do user layering and demand digger. Now we center around different format. We do verification. Talking video is a format, and MV is another format. We believe every format deserves larger potential. For those who did not perform well, we will center around different metrics to evaluation. For example, we merge or we cancel to ensure limited resources are spent on good-performing products in the future. We will not extend our boundaries endlessly. We will not do things which do not relate to our field or professions.
The AI content creation tools, it will naturally extend to distribution because all content needs advertising. Many content could be accumulated as users' assets or IP. This is just a flywheel of content. Centered around this flywheel, we can verify or check the authenticity of many assets or IP. I believe, or we consider, compared with the pure application in the past, we can better satisfy users' needs. The content created could be seen, it could bring revenue to users. This is actually common demands of our users and how to accumulate long-term valuable assets. Based on insights above, we form our new strategy forward. Thank you.
About how to make ARPPU value higher, core lies in finding user's quality demands. They would use this product more frequently. Of course, choosing the path is also important as well. For creating MV in the era without AI, it would be expensive, around CNY 50,000- CNY 100,000 in the past, but now with AI, it is much cheaper. If we pick products from these fields, the ARPPU value would naturally be higher.
Thank you for your question. Next, lady from the virtual.
Just a minute, please. Hi, management. Lydia from Morgan Stanley. I heard that these products moves from creation to distribution to accumulation is really innovative, a new perspective to see Meitu. But in the past, people are positioning Meitu as a light asset tool-type app company. But I heard that you will be exploring into distribution. We do not know if you are moving into operation. Please, management, elaborate on that, including the impact of future business model and management model as well as finance. Thank you.
Okay, I will start on this. First, we believe that distribution, we have mentioned, is really a shared demand of the industry. We can see that whether the app itself or others, they all need distribution, including game, media, and music. All industries, the value distribution.
In this field, we worked on it previously too, but rather on the app itself. Now, aligning with the creator's needs, we are expanding distribution to content. From team building, we can say that based on the current distribution team, we are expanding that. We are building on that. It is not starting from the ground because we have already been doing it.
Yeah. Okay.
I would like to add a point, that is, distribution is more like better meeting creators' demand. In the past, Kaipai and MVLAND, these creation tools, after completing the content, user would naturally want to distribute. But in the past, we only have the first step ready, and users would have to distribute for themselves or use third-party channels to better attract traffic or distribute into social media and e-commerce and live streaming. Distribution is more likely based on creation. It is a back chain service.
It's not brand new. It's really just an add-on to the current product. After creation and distribution, the creators will eventually on different platforms and in its own brand, they can have their own persona, their assets, their licensing rights. You can understand that as we are serving the more diversive needs of the current consumers. This is a new cycle. We have a higher value for the users.
Next question on the floor. Analyst Vicky.
About the distribution of content and the accumulation of assets, can you share more about what are our advantages and what challenges did we face? Second question, the MAU overseas, this increase in speed has slowed down. What trends or differences or changes did the management observe?
First, I answer about our advantage and the challenge. For distribution, one major advantage is that we have our own channel. We have our app matrix, Meitu app, BeautyCam. They play a very good role in distribution. For example, many good IPs, they've turned to Meitu's app to do their distribution. We can say that Meitu app can verify many IP or creation assets, whether it could be loved by users and from users, can it be more popular among users compared with other companies? This is our advantage.
We have our own channel or source. Next, when it comes to challenge or shortcoming, how to build external sources or channels, other social media, online video platform, or e-commerce platform. They are, in broader sense, distributing channel, how we can buy view metrics. This is not pure disadvantage, but we just didn't develop relatively much for distributing platform. This would also be a major goal for us. For products overseas, we now focus more on Higgsfield in U.S. Its ARR was $500 million. Now it reached $700 million. You could take it as there are so many B- end users and studios, advertising, and creating companies, they would use this to do their advertising video on social media. Because we cannot 100% copy its mode or model, but we can learn from it how we can build distribution channel to sell other products.
Because one of its logic, the products needs to be advertised on social media. You would find MCN and other companies, you give them or they give you a budget, they would find the 1,000 or 100 K OL to do advertising. Among them, 10 might go viral, but this is traditional. Now we focus more on the efficiency. If you already have a bunch of creator users, if they use your tool, they will naturally publish it. If you let them just add your tag, if the content they make goes viral, they can submit this link to our creator platform. We would analyze the metrics of this video or work. We would also give them kind of a revenue or boost. This is different from what we do in the past on social media.
In the past, we do not know whether it would go viral, but now we make sure we pay this money to those viral videos, for we can take advantage of the long tail creators, but it does not mean that it has a prerequisite. Not all videos can be popular or viral on the platform, but we need to find those ones. For MVLAND, RoboNeo or Kaipai, it is naturally suitable for social media. So we can rely on good, high-quality creators to achieve better advertising and distribution.
Thank you. Now we move on to next question. We have a question from [Common Finance]. Ms. Chen Xi from China Securities. We are unmuting you right now. You are unmuted. You can start questioning any minute.
Next question, Chen Xi from China Securities. You can start any minute.
Due to technical issues, we will move on to the next question.
Okay, now we move on to the next session. Thank you for your questions and answers from management. If you have other inquiries about the performance, you can contact the IR team. Next, we move to media Q&A session. This session will not be interpreted. You can leave the Zoom meeting room now. Thank you again for your participation.