During this Spring Festival holiday, a song called "Ascending Spring Mountain" became popular. The public enthusiastically learned the various techniques for seizing the spotlight, known as "Spring Mountain Studies," during the festive period.
As the holiday comes to an end, the mobile phone industry is vying for the spotlight this year. So, where is the biggest opportunity for smartphones this year? The answer is clear: AI large models. In the second half of last year, mobile phone manufacturers successively launched on-device large models. Shortly after the Spring Festival, OPPO announced a new AI strategy, declaring its entry into the AI era. Meizu announced that it will no longer produce "traditional smartphones" and will fully embrace the AI era. Clearly, large models are the "Spring Mountain" of the mobile phone industry.
However, according to the knowledge of "Spring Mountain Studies," it's not enough to just follow everyone up the mountain; one must go further to secure the spotlight. Among the numerous combinations of large models and mobile phones, there is one capability that stands out, the "true unparalleled Spring Mountain" – large models combined with mobile phone imaging.
It is said that during this Spring Festival, photo studios in first-tier cities have transformed their business models. Instead of taking traditional portraits, customers now take AIGC photos, and the studios have transformed into AI prompt engineers, using various AIGC platforms to meet users' diverse and unconventional needs.
Comparing the photos upon returning, it's not about which one has better lighting or looks more natural, but rather, which prompt words were used for a particular photo, and which one exhibits stronger model generalization…
In short, the AI trend is sweeping through our visual lives, from photo studios and portrait studios to mini-programs and apps.
But here's the question: do these strong AIGC imaging demands really have to be fulfilled in offline stores? Why can't they be easily accomplished on mobile phones? Recently, new mobile phone models have been equipped with on-device large models with billions of parameters. What are they doing with all that capability?
In fact, the large models in mobile phones are definitely not idle. Currently, for mobile phones embracing AI large models, the situation is that large models are large models, and imaging is imaging, but this split situation is bound to change soon.
The "fusion track of large models and mobile imaging" is the clearest trend in the 2024 smartphone industry, and it is a strategic "Spring Mountain" that smartphone manufacturers are eager to climb and must climb, and once they do, they cannot come down from.
Large models + mobile imaging, how do we ascend this "Spring Mountain"? Let's delve into the "Spring Mountain Studies" to understand this issue.
Six years later, AI imaging ascends the mountain again
First and foremost, we need to be clear that AI + imaging is by no means a new concept, and many people have already become accustomed to the combination of AI and imaging.
This concept began to receive attention from the industry as early as 2018. In April of that year, Huawei released the P20 smartphone. This phone, utilizing the NPU on the Kirin chip, incorporated AI functionality into photography for the first time, capable of recognizing 19 different scenes, covering over 500 recognition targets, including pets, portraits, landscapes, and food. After recognizing these objects, it would automatically adjust the photography mode and parameters.
Subsequently, this AI photography mode became extremely popular, becoming the most heated technological upgrade for smartphones at the time, and gradually gained acceptance across the industry. With several years of development, AI imaging has become increasingly complex, capable of recognizing a wider range of objects, and has added abilities such as dynamic capture and glare removal, becoming one of the fundamental capabilities of mobile imaging systems.
However, at this stage, the capabilities of AI photography still have significant limitations. Its impact on images is primarily focused on "beautification," rather than "modification" and "generation." We have also spoken with developers in the AI photography field, and they are eager to utilize the AI capabilities of mobile phones to create more creative applications, but on-device computing power and model capabilities are the main limitations.
With the explosion of AI large models, the bottleneck of model capabilities has been broken. With the support of large models, users can issue complex commands to the imaging system, and the imaging system can better understand user interaction logic and intent. In terms of capabilities, large models can help achieve high-precision image element replacement, and even incorporate AI-generated images.
It could be said that the imaging capabilities that users initially imagined when they first heard about AI imaging concepts are only now becoming possible with large model imaging.
With the integration of large model capabilities into mobile phones, AI imaging has finally made a crucial leap, fulfilling the promises made long ago.
Large models are the essential "mountain" for AI imaging.
On the Spring Mountain, what is the scenery like?
Before the Spring Festival had even ended, the world experienced visual shock from Sora. There is a saying in the AI industry: language models are for making a splash, while visual models make money. Machine vision capabilities are the fastest and most effective way for users to experience the charm of AI.
For a long time, mobile imaging has been continuously evolving, but users have always been limited to taking pictures. With the addition of AI large model capabilities, users can now effortlessly modify images and combine AIGC images with images captured on their phones. The combination of AI large models, mobile AI computing capabilities, and mobile imaging systems has greatly expanded the boundaries of mobile imaging, inheriting the technological advancements and supply chain layouts of mobile manufacturers over the years, while also gaining new growth opportunities.
At this stage, this track has rapidly unfolded. For example, Samsung has enabled users to move objects within images, automatically fill in gaps, and generate new images through the Picture Assistant feature, providing greater compositional freedom for mobile imaging.
On the other hand, the OPPO Find X7 series has implemented AIGC elimination functionality through AI large models. As seen in recent advertisements, users can eliminate unwanted individuals from group photos taken during the Spring Festival and fill in the background using AIGC. Currently, the Find X7 can support the separate extraction of up to six subjects. In addition, OPPO has also updated the AI Ultra Clear Portrait feature, which intelligently recognizes and enhances the clarity of faces in group photos.
It is foreseeable that we will see a large number of image functions implemented based on AI large models, such as AI cutouts, AI replacements, and AI image expansion. Overall, the combination of large models and mobile imaging will present three major development trends:
Integration of AIGC content with captured content. The AIGC's "text-to-image" platform has rapidly gained favor with users over the past year. The critical battle in the AI smartphone track is to integrate this capability with the inherent imaging capabilities of mobile phones.
AI capabilities move up from the application side and integrate with the mobile phone's imaging system. Currently, the visual capabilities brought by AI large models are primarily standalone software applications. In the future, mobile manufacturers will move these capabilities to the system level, making them a differentiating selling point of the product itself.
Mobile imaging capabilities can be updated over-the-air (OTA). With the addition of AI large models, the imaging capabilities of mobile phones have become upgradable and iterative software to a certain extent. This allows the system-level capabilities of mobile phones to be continuously updated and operated, representing a new change brought to mobile phones by AI large models.
Overall, the combination of large models and mobile imaging provides a very abundant space for creative expression and continuous possibilities. To seize this opportunity objectively, it will bring about a new round of technological competition among mobile phone manufacturers.
That mountain is the next strategic high ground
At this stage, no manufacturer has explicitly put forward the concept of "large models + imaging," but this concept has already landed under various names and should not be far off.
It is important to note that while deploying large models on the device and developing creative AI imaging capabilities are relatively easy, turning AI large models + imaging into a long-term track and forming a brand's user mental support point will require manufacturers to invest significant effort to launch a new competition in the smartphone industry.
The reason for this is that large models + mobile imaging is a comprehensive test in every sense. It requires the collaboration of hardware-side computing power, the support of imaging systems, algorithm support from the model side, as well as creative application development and AIGC-specific visual aesthetics. From practical to virtual, from underlying hardware to top-level applications, the competition of large models + mobile imaging covers almost every level of the mobile phone industry.
Focusing on seizing the strategic high ground of large models + imaging requires mobile phone manufacturers to concentrate their efforts in three areas:
1. Update AI infrastructure. This infrastructure includes both AI computing power and AI algorithms. It involves chip capabilities, edge-cloud collaboration capabilities, and basic algorithm capabilities. The demand for updating the AI infrastructure of mobile phones will also drive a new round of reshuffling in the industry chain.
2. Accurate grasp of AI applications. The possibilities that AI large models can bring to mobile imaging are not too few, but too many. How to provide users with the most accurate and appealing AI imaging applications under limited computing power will be the first challenge faced by mobile phone manufacturers.
3. Development of AI aesthetic capabilities. Many people have recently noticed that the Sora team has specifically recruited artistic talents. Art and aesthetic capabilities are very important in the AIGC era. As the industry develops, algorithm capabilities will converge, while differences in aesthetic capabilities will be exposed. In the past, mobile phone manufacturers focused more on design capabilities, which are somewhat different from artistic expression and aesthetic judgment. Building a new aesthetic capability specific to AI smartphones is a completely new competition.
Finally, here's a somewhat counterintuitive judgment: the "Spring Mountain" of large models + mobile imaging is actually quite easy to climb. Its threshold is far from being as exaggerated as the mobile phone manufacturers claim. However, if you have recently studied the "Spring Mountain Studies" material diligently, you will know that the real difficulty is not in climbing the mountain, but in not wanting to come down from it.
Turning large models + mobile imaging from an annual gimmick into a long-term track for multi-year development, and even turning it into an anchor point that reshapes the industry landscape, that is the real test.
The good news is that having a "Spring Mountain" to climb at least proves one thing: the solid ice enveloping the mobile phone industry is cracking and melting under the warm breeze of technology. Whether to bid farewell to winter and embrace spring is in the hands of the practitioners.
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。
