Image source: Generated by Unbounded AI

It’s not that Jellycat is unaffordable, but AI generation is more cost-effective

Recently, Meituan’s “Miao Brush AI” launched a gameplay that makes everything Jellycat-like. Just upload a photo, and AI can automatically recognize the image and turn the subject of the photo into a Jellycat doll, along with the doll's name, price and an abstract care guide.

Snow White is transformed into "the strongest worker in the fairy tale world. She makes money in the castle to buy bows during the day, and goes to the forest to teach math to small animals at night." Its care guide is "Never let her know that the pumpkin carriage is actually It's rented." Naruto's maintenance guide is to practice the "Plush Clone Technique" with it every day, which is outrageous with a hint of reasonableness.

The classic small eyes and smiling mouth quite restore the characteristics of Jellycat and are very cute. Many netizens said that "the design is more innovative than Jellycat's" and "it feels like the Jellycat designer is going to lose his job."

In fact, before Miaoshu AI launched this gameplay, Xiaohongshu had a trend of using AI to generate Jellycat pseudo-co-branding.

However, most tutorial posts require the use of AI image generation models such as Midjourney, Dream AI, etc. to generate input prompt words. It is not as good as Miaobhu AI has low threshold for use, convenient operation and personalization.

In addition to Jellycat, which can be used for everything, the Miaoshu AI applet/App also has "Crazy Menu", "Three Kingdoms Brush", "Magic Item Collection Manual", as well as one-click transformation into cube people, minions, Harry Potter wizards and other picture-based gameplay are fresh and interesting both in terms of the creativity itself and the generated effects.

Meituan’s exploration and actions in AI have always been relatively secretive. Miaoshu AI is an AI picture generation application produced by the AI ​​team outside Meituan Lightyear. In early November, it hit the screen in the AI ​​circle group chat.

Today we will start from the perspective of Miaoshu AI gameplay design and product communication, and see what is "wonderful" about Miaoshu? What are the possibilities for subsequent commercialization exploration? Let’s take another look at Meituan’s AI layout.

What’s so wonderful about using AI?

After some experience, Miaoshu AI gave us the biggest feeling: it understands Generation Z very well and is good at playing with abstractions.

Emphasis on "exploration" in interaction design, slide up and down to try different templates. And most of the templates revolve around popular IPs, such as Jellycat, Harry Potter Academy Identity, Genshin Impact, Minions, etc. According to this logical inference, there are still many content directions and gameplay methods that can be tried in subsequent Miao Brush.

More importantly, the ease of use of uploading photos to get the same style and social sharing is suitable for any user to get started quickly, and the experience is very smooth and smooth.

I first uploaded a picture of a pillow at home, which I thought was true.The object is already very cute, but I didn’t expect that the same model as Jellycat would be twice as cute...

The care guide below is also very distinctive. , does not have too much of an AI flavor - "an Apple guy suffering from social fear", this is indeed his own portrayal.

In addition to the popular Jellycat template, Miaoshu has also recently launched a new "Crazy Menu", which can combine elements from any two pictures to generate a double meal with an ingredient list and dish introduction. No, the CP fans who are cooking are here just because they smell it.

I uploaded a picture of a self-deprecating bear and a picture of a red Queen Mother tote bag, and got a bear rice ball toast. From the "cute bear and luxury shopping bag" in the text introduction, we can see that the picture recognition is very accurate. "After eating it, it can improve your fashion taste, and the side effect is that you will become more fond of shopping." It is also very unique.

The gameplay of "Three Kingdoms Brush" can convert uploaded pictures into Three Kingdoms killing cards, and there are SSR, SR, and S card slots. After all, drawing cards is a big deal. For example, uploading a photo of Genshin Paimon and sending it directly into the soul SSR. Is this the power of "emergency food"?

The "Magic Item Collection Manual" method can output a very interesting introduction to any uploaded item. Whether it is a trash can, a camera, or an abstract emoticon package, does it feel a bit familiar? Yes, this is similar to the popular "New Interpretation of Chinese" prompt. They both interpret an ordinary object from a new perspective.

"The disappearing coins eventually turned into digital wandering souls", "Twin Soul Washing Machines"... This idea, if Even an AI that is fed abstraction after abstraction will not be able to write. Since staying up late and memorizing tricks can't compare to talented players, let's find AI as a foreign aid.

“Gallery” is a newly launched social square on Miaoshu. Here you can see the works shared by others and create the same style with one click.

The Miaoshu team introduced that Miaoshu is positioned as a tool that provides novel, fun, and absurd AI visual experience tools. It hopes to make the daily unchanging life scenes interesting and colorful, which is a point that faces boredom. Tiny resistance.

According to the AI ​​New List, Miaoshu is their attempt to explore multi-modality and workflow. They are firmly optimistic about ComfyUI and believe that it will have many interesting things in the direction of multi-modality. Possibilities arise. Currently, Miaoshu AI provides 14 templates, and they are recruiting more and more interesting ComfyUI creators to create more new gameplays with them.

Behind Meituan’s diversified AI layout, major manufacturers are also confused?

Since the emergence of AI image generation models, stylized fine-tuning models have been very popular, including Flux's realistic style and Recraft's dream core.It has a film style, but Miaoshu’s product design goes one step further, adding copywriting and layout on top of raw pictures, and using the ComfyUI workflow to provide a richer gameplay experience.

Currently, Miaoshu AI is still in the limited-time free trial stage, and a certain amount of generation can be obtained through social sharing.

As for how to convert traffic into commercial value? A common idea is to encourage users to share through product incentives and operational activities to social media such as Douyin, Xiaohongshu, and Weibo to increase product visibility. The popular AI video special effects are a typical example.

However, considering that Miaoshu AI is backed by Meituan, it may be more relaxed in commercial exploration.

In terms of exploring the application level of large-scale AI models, Meituan’s steps are diverse but not radical.

As early as when the first quarter financial report was released last year, Meituan founder Wang Xing stated that the company had established an internal team to conduct research and development on large models and application levels. But while Baidu, Alibaba, and Byte are all promoting their large model products with great fanfare, Meituan’s large model progress has always been kept secret.

When it comes to Meituan’s AI layout, perhaps the one who is most aware of it is the acquisition of Wang Huiwen for 2 billion yuan. In addition, Meituan has also invested in two large model companies, Dark Side of the Moon and Zhipu AI, and has invested in Yushu Technology and Galaxy General Motors. It also has investments in robotics, unmanned delivery, autonomous driving, chips and other fields.

For its core local life business, according to Tech Planet’s report in April this year, Meituan has developed a C-side large-model dialogue product for Meituan takeout called “Miru”. It aims to improve users’ ordering experience through intelligent recommendations and chat functions. Soon, Meituan began testing the AI ​​assistant service "Ask Xiaobao". This is a takeaway intelligent assistant product based on the Raccoon model, which can recommend restaurants to users, provide gift-giving suggestions, etc.

After being acquired, Lightyear Waiwai was responsible for exploring some independent AI applications within Meituan. Currently, it has launched multiple products in areas such as AI social networking, AI imaging, and AI + children, but unfortunately What's interesting is that it didn't cause much trouble.

The AI ​​social direction is mainly the AI ​​companion chat product WOW, which allows users to communicate with freely customized AI characters through voice and text, and also includes a variety of AI interactive experiences such as interactive stories and reading.

In the field of AI image generation, in addition to Miaoshu AI, which was launched in September this year, in February this year, Meituan briefly launched the AI ​​photo gallery function within its application. It is placed discreetly in the carousel display column of the "Wedding Photography" page. The functional design is very similar to the Miaoya Camera. Users upload multiple photos from different angles, and AI automatically generates a set of AI portrait photos.

On the AI+children's track, Meituan Lightyear has launched two products. One is Dodoboo, an AI-driven children's painting product for overseas, which can optimize and generate children's simple graffiti. A beautiful pictureComplete picture. The other is the AI ​​voice interactive APP "Qiaoyu Leban" launched in cooperation with Xiaotiantian Watch for children. The Xiaotianz10 watch equipped with this APP was launched in May this year.

According to Wang Xing’s previous statement: “In the first half of technological change, because the risks are very high, small teams are needed to explore. But in the second half, the dividends become smaller, and integration becomes a release dividend. "Meituan insiders once revealed: "Meituan has to do something. It’s not just about large models, but it also needs to be integrated with its own business. It’s in the vertical field of large models. There are too many directions, and they are all still being explored. “The industry is gradually shifting its focus from breakthroughs in large model capabilities to the application of large model capabilities. When the time comes, Meituan will focus more on exploring AI applications.