Source: Baijing Chu Hai

Author: Amber

Editor: Zhi Ting

Image source: Generated by Wujie AI

Some time ago, data.ai released a list to celebrate the one-year anniversary of the launch of ChatGPT (desktop version released in November 2022), listing the top 10 AIGC applications with the highest global download volume since the release of the mobile version of ChatGPT (May 2023).

In the list, most are chatbots similar to ChatGPT, with only 3 apps classified as AI art generators, which are the AI image generation apps we are familiar with. In the stereotype, whether it's cartoon faces or the popular Yearbook app from a few days ago, most of the apps that generate stylized images based on AI are quite similar. In terms of download rankings, they are straightforward.

However, after carefully examining the data, we found that the seemingly similar AI image generation apps seem to have found their audience in the U.S. In terms of active user volume, the best-performing AI Mirror in the U.S. market had a daily active user (DAU) count that was higher than that of "Moonlight" FaceApp (this app's monthly revenue in the U.S. is close to $3 million). Although there is still a significant gap in revenue generation, this DAU performance deviates significantly from our own perception that "many people download it, play with it for a few days, and then leave it to gather dust."

During the Google Overseas Entrepreneurship Accelerator Showcase held at the Beijing office, we met Steven Guo Xuan, the founder and CEO of Polyverse, the company behind AI Mirror, and Queena Qiu Zijun, the co-founder and CGO. Before meeting Steven and Queena, we obtained some dimensional data about AI Mirror through third-party platforms. If we say that the stereotype of a wave of AI image generation apps has been broken, there are still several questions that we are very curious about.

AI image generation is not a rigid demand, so who is supporting the hundreds of thousands of DAUs in the U.S.? Where does the competitiveness of AI image generation apps lie? How do practitioners in the AI value chain view the development of the entire industry?

From gaming to AI, we have always regarded AI Mirror as a "toy"

After graduating from university in 2016, Steven worked at Tencent for a few years on MMORPGs. "At that time, I felt that the overseas gaming market had a lot of opportunities, so we decided to start a business. At first, we did hyper-casual games, but later, due to changes in the traffic environment, we shifted to casual games."

In 2022, a new opportunity arose. "I have a technical background, so I pay a lot of attention to new technologies. In the second half of 2022, after noticing AI, we turned our company's employees into superhero images and created some funny emojis. Everyone found it very interesting. We also evaluated it and found that we could do it ourselves, and we did it quite well. After researching the products on the market, we believed that there was still room for improvement in terms of visual appeal, user element retention, and the simplicity of user experience. So, in the second half of 2022, the demo of AI Mirror was ready, and after a long period of preparation, the product was officially launched in March of this year."

The most surprising thing about AI Mirror is that, even after reducing advertising, it still maintains a relatively good DAU curve. Unlike photo editing, which is a "rigid demand," AI image generation apps have maintained a stable audience.

"For AI Mirror, its positioning is the same as our original intention, which is a toy that provides a novel experience. After the novelty period, there is usually a lot of churn. It indeed does not have the same rigid demand as photo editing, but we found that this toy matches the needs beyond photo editing and sharing. On our platform, there is a group of users, somewhat similar to creators or small business owners, who use AI Mirror for creation. The stickiness of these users is reflected in their frequency and duration of use, which is even higher than that of photo editing applications.

For example, some users make greeting cards in our product. The stylized images created by AI Mirror add a certain level of fun and richness to the original images, while also retaining many details of the original photos, making the scenes more 'emotional' and expressive."

After participating in this phase of the Google Overseas Entrepreneurship Accelerator, the Polyverse team, through contact and communication with mentors, believes that AI mainly has two directions at the application level: one is how AI can innovate traditional/offline scenes, such as the comparison between Meitu camera and professional photography; and the other direction is to use AI to create incremental value, rather than thinking from a replacement perspective. "Just like when mobile games came out, they expanded the gaming market instead of completely squeezing out the PC and console markets. In the field of images, AI image generation provides stylized value, unleashing user creativity, and intelligent photo editing to a certain extent replaces traditional photo editing, which is also a huge opportunity, but it is no longer the only solution," Steven added.

After contacting the team, we further observed the entire image editing market and found that, in addition to AI Mirror, there are actually a few AI image generation products with very good performance in terms of DAU.

AI image generation apps launched around 2022 have already reached a level of DAU in the U.S. market that can compete with top photo editing apps | Baijing Chu Hai based on data from DianDian (blue background represents traditional photo editing, yellow background represents AI image generation)

In addition to possibly meeting the needs of users who want to unleash their creativity with the help of AI image generation apps, newly launched AI image apps have targeted a wider range of users with a lower usage threshold, subscription-based monetization, and advertising.

Traditional top apps have established a stable user base and formed certain competitive barriers, so their paid features are very strict, with only a few free items available. Although they have also embraced AI, most of them have set up paid features. However, newcomers are more willing to cover more users with a free experience, which is also a reason why this wave of new forces can compete with traditional apps in terms of active user volume.

AI image generation apps have adopted different commercialization models for avatar generation

The sustained value and competitiveness of the "toy"

By entering the AI image generation market and providing a stylized image generation experience, AI Mirror, like many AI image generation apps, has achieved good traffic in the initial stage.

The download volume of AI Mirror experienced explosive growth twice, once after its launch and again in May. "The burst of growth at the end of May was mostly organic, coming from users' self-propagation on social platforms. We found a very interesting phenomenon, which is that users will spontaneously spread on a regional level, based on common cultural awareness, from one market to another. So, when we felt we could actively grow, we would focus on certain core markets, seizing the momentum of social media propagation to radiate to surrounding areas, such as Indonesia. In the Southeast Asian region, Indonesian users are very curious, and the download volume is very high, so it can be used as a pilot," Queena summarized.

Although the product growth process also matched some sticky users, fundamentally, as a "toy" AI image app, there will indeed be user churn and rapid decay. In response to this, Steven shared his two thoughts:

First, seize and amplify the niche users mentioned above.

AI Mirror was initially positioned as a toy, targeting user sharing needs, and with commercial considerations, this led us to balance the speed of image generation, usage scenarios, and so on. These scenarios may not be suitable for the B-side creation scenarios observed by our sticky users. In the future, we will iterate quickly based on the needs of this wave of users.

Second, broaden the scenarios.

User behavior corresponds to the app's scenarios and functions. Currently, most users still create good-looking avatars and then share them on social media. We hope to broaden the sharing scenarios, allowing them to have more content to express, and another aspect is whether we can upgrade the technology to extend to the video scenario."

During the execution of this dialogue project, we saw that AI Mirror has launched the Magic Brush feature, which uses AI text commands to replace certain elements in photos, AI Video functionality, multi-person image stylization functionality suitable for overseas holiday seasons, and templates that make it easy for users to send Christmas cards.

Continuously enriching user scenarios has allowed the app to maintain a good user base even after reducing advertising efforts, but it also places significant demands on computing resources.

"The computing cost of AI image generation is a huge burden, and we have done a lot of work at the computational inference level to reduce costs," Steven said. "At the same time, in terms of growth, we pay more attention to word-of-mouth spread among users to build brand awareness, thereby reducing the overall user acquisition cost. On the commercialization front, we launched the business model on the first day, focusing on subscription and advertising monetization. We have only raised seed funding before, and the scale is not large. We hope to rely on our own efforts to accumulate and then invest more resources when needed."

In conclusion

AI experienced a venture capital boom in 2023, and many projects were overturned in each iteration of GPT. Behind the huge opportunities is also a test of the understanding of the AI entrepreneurship track.

Creating application-layer AI products is seen by most people as "shell-based," fundamentally still a traffic business. However, from the current situation, as users continue to interact with the app, AI products originally designed to provide a novel experience are beginning to be explored by teams for deeper value. The accumulation of data and algorithm optimization also allows products to gradually establish some competitive barriers.

In today's business environment, the exploration and definition of product value, algorithm optimization, growth strategies, and commercialization models all affect whether an AI image generation app can stand and succeed in the market. AI technological innovation provides an opportunity for capable entrepreneurial teams to enter the already crowded "old track," and their exploration may yield different results in 2024.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。