A：The progress of our cooperation with Alibaba is basically in line with expectations. As we mentioned earlier, we have gone through 17 years of development, and our products have gradually grown and iterated over a relatively long period of time. So our cooperation with Alibaba will also go through this process, and it is not expected to generate significant revenue or user growth immediately after the announcement of the partnership. I believe that it takes time to build trust between the two teams, and to find complementary positions. Of course, as I mentioned earlier, we have some basic cooperation initiatives currently underway, so overall it is quite positive.

A：We maintain cooperative complementary relationships with all major model manufacturers. Rob's new product builds model containers, integrates various basic models including Ali's Tongyi Qianwen and Wanxiang, as well as vertically trained models, and flexibly schedules them based on effectiveness and cost. We have established partnerships with many model manufacturers, currently achieving good results, allowing us to focus on improving our products and effectiveness. At the same time, we also see that some technically-oriented model companies hope to enter the application field, not every company can successfully transition, and the relationship between the two is more complementary than competitive.

A：In terms of marketing budget, we generally maintain it at around 16% of revenue from image and design products plus advertising income, with slight adjustments. In the long run, as subscription penetration rates increase for Chinese lifestyle products, sales expenses will decrease, allowing more resources to be allocated to promoting overseas products. As for research and development expenses, we no longer conduct basic model training, mainly focusing on personnel expenses and small vertical model expenses. We expect a growth of less than 15% in the coming years, not by deliberately increasing personnel, but by optimizing AI utilization to drive innovation.

A：Due to the rush to launch, Robo Neil first introduced a daily card payment mechanism, and will soon launch a complete subscription or single purchase payment mechanism. Currently, the number of monthly active users exceeds one million. Although revenue is still in the early stages and has not been disclosed separately, the results are very encouraging. A product that does not rely on internal traffic and has zero marketing budget can achieve over a million monthly active users in a short period of time.

A：The video generation capability has been widely used in Meitu's existing products, such as wink and kaipai. Video products are growing rapidly, with wink already reaching over 30 million monthly active users, far exceeding the growth rate of image products, and with huge potential for monetization and productivity-related applications. Video investments can also benefit image products indirectly by being downwardly compatible with images. The video generation technology model is based on the Ali Tongyi Wanxiang video large model, and multiple vertical models are trained on top of it, some of which are trained by us to improve the effect. In the future, each productivity product is likely to have expert models, such as training and providing services for video directors from an expert perspective.

A：Overseas development is divided into two main lines: the lifestyle line and the productivity line. The lifestyle line closely follows the demand and trends of social media content production, seizing opportunities through collaborating with Key Opinion Leaders (KOLs) for word-of-mouth marketing and continuously increasing investment. The productivity line focuses on vertical scenarios, addressing pain points and aiming for precision. A recent example is providing brand visual production products for overseas offline stores, extending services from small stores to small companies, combining online and offline channels. In different regions, despite their unique characteristics, the core remains capturing pain points in vertical fields.

A：We chose Alibaba Cloud's Qianwen model because in the competitive landscape of large models, we are basically unable to cooperate with closed-source models and can only rely on API calls. Among open-source models, such as the deep c model that became popular during this year's Spring Festival, lacks the ability to understand multiple modalities compared to Alibaba's Qianwen and Wanxiang, especially in image and video processing capabilities. It is difficult for us to use directly for our application scenarios. In addition, Alibaba has made huge investments in models, possesses powerful multimodal understanding capabilities, and has long been at the top of the model leaderboard. Even without a capital strategic cooperation, we might still choose them, but cooperation can allow us to more deeply meet our needs, such as targeted training and early access to the latest versions, all of which will translate into the competitiveness of our products.

A：We continuously explore vertical scenes and enhance product competitiveness, returning to the product itself, allowing good products to grow naturally. The product has made significant progress, but efforts are still needed to create a world-class product. As for the payment model, there may be more innovations in the future, but specific details have not been elaborated.

A：We believe that there is growth potential for domestic users. With the emergence of third-generation product forms (such as language interaction based on large models, multimodal understanding, etc.), the industry will usher in tremendous changes and the dividends of the times. Embracing this trend wholeheartedly is expected to bring incremental space to the Chinese and even global markets. There is still growth space domestically, particularly in the fields of images and design. The global arena has a lot of room for development in this field, with rapid growth in productivity, as shown by examples like Kuaipai.

A：Meitu currently does not develop video basic models, but instead uses Alibaba's Wanshixiang super model as a foundation to develop vertical models to meet various functional and effect requirements. Unlike companies such as Kuaishou and ByteDance, Meitu does not independently research basic models, but will cooperate with other model capability providers such as Jiemeng and Koning.

A：The difference of Meitu lies in its focus on the track of imaging, design, and aesthetics. Meitu has over a hundred internal designers focusing on researching new trends in effect design. Its core competitiveness includes large-scale engineering investment, deep insight into aesthetics, and engineering optimization capabilities. The challenge Meitu needs to overcome is a profound understanding of the cultural needs of different countries around the world, especially in regions like Europe and America.

A：The company will continue to increase its investment in image and design products, with a lot of growth potential in this sector. On the one hand, there is still a lot of room for improvement in the subscription penetration rate in daily life scenarios; on the other hand, productivity products have a higher subscription penetration rate due to their direct impact on user's expenses. Although no specific number guidance was provided, it is evident that the growth potential in this area is huge.

A：Building AI agent products is to seize the opportunity of the third-generation product form and to prevent missing the market opportunity. The product design emphasizes differentiation, such as focusing on vertical fields of imaging and design positioning. User acquisition mainly relies on word-of-mouth driven by product strength, rather than traditional promotion strategies, to test the market performance of new products.

A：We will enhance the productivity attributes of Rope Neil, focusing on developing the desktop web version, refining the common needs of global users, and introducing a product capability that everyone can use.

A：The model container strategy focuses more on improving product performance and user satisfaction, rather than directly reducing costs and increasing efficiency. It adopts a horse race mechanism to select the optimal solution based on model performance, enhancing competitiveness and user satisfaction, while also learning from external models to enhance its own technical capabilities.

A：Meitu expects that over half of its income and profits will come from productivity tools, and will continue to launch competitive products. In addition to existing designers and photographers, there are also products such as VWHEE that have transitioned to entertainment material design. In the future, Meitu will continue to launch more products and serve global users well, especially users in Europe and America.

A：The popularization of features such as AI fill light is attributed to users' spontaneous sharing and imitation. Meitu considers growth factors when launching new features, preferring to find opportunities for amplification from user-generated content with good word-of-mouth and high recognition on social media, rather than relying solely on internal research. Now the company is giving users the creative control, discovering and utilizing growth points from user-generated content and social media dissemination.

A：Meitu considers itself as an AI application company, because AI has a high penetration rate, which has been continuously embraced during the development process. Meitu divides its products into native AI applications (such as Roborock) and non-native AI applications (such as Meitu Xiuxiu), both complementing each other. Regarding the relationship between traditional editors and AI-generated content, Meitu believes that they should complement each other, rather than directly transforming existing products. Meitu encourages internal challenges to existing products, trying to break away from traditional frameworks, and create a completely new form of AI applications and technical architecture to achieve better functional innovation.