Image source: Generated by Unbounded AI

At the "Technology Spring Festival Gala" CES 2025 two days ago, in addition to the AI ​​glasses booth being crowded, there was also a long queue in front of the AI ​​headset booth.

With the explosion of the AI ​​industry, the battle for AI entrance has become a major wave in the past two years. AI headphones are a market as big as AI glasses.

Currently on the market, mobile phone manufacturers, audio brands, major Internet manufacturers and technology companies are all making AI headsets. According to incomplete statistics from "Fix Focus One", there are no less than 14 headphone brands on the market that have stable sales and truly have AI functions.

There are two major premises behind the popularity of AI headsets.

The first is the popularity of TWS (True Wireless Stereo) headphones, which are true wireless stereo headphones. In 2016, Apple released AirPods, and wired headphones gradually exited the market. The increase in outdoor activities in the past two years has intensified the penetration of wireless headphones.

Canalys data shows that in the third quarter of 2024, TWS headset shipments reached 92.3 million units, a year-on-year increase of 15%. Xiaomi and Huawei have both achieved annual growth rates of over 50% in the global market. However, the industry is currently relatively fragmented. Apple, which ranks first, only has a market share of 21%. With the addition of AI, market growth and industry structure may change.

The second is that the large language model has developed to the GPT-4o stage, and the quality of voice interaction has become more mature. Compared with the smart speaker era, you have to wake up first before talking every time, and the speaker either does not understand or cannot answer, and the usability Greatly improved.

Based on these two points, in the second half of 2024, many manufacturers will release new AI headsets. Industry insiders predict that the AI ​​headset market will usher in explosive growth in 2025. By linking headsets to smartphones, the real Siri era is not far away from us.

The AI ​​​​of headphones is unstoppable. Who can reap the biggest dividends? Who is the biggest winner?

Three major schools of AI headphones

Earphones are basically wireless After that, consumers’ usage time of headphones increased significantly, and the usage scenarios also extended to meetings, sports, etc. Starting in 2024, the headset industry will once again enter the AI ​​process, adding AI voice interaction functions to the headsets.

According to "Focus One", the companies currently entering the AI ​​headset industry can be divided into three major groups, and their entry purposes and product directions are also different.

The first group is traditional mobile phone manufacturers, such as Huawei, Xiaomi, Samsung, etc. Taking Huawei as an example, users wake up the "Xiaoyi" intelligent body through AI headphones, and then query routes, navigate, and perform face-to-face translation ( What others say is directly translated and sent back to your headset) and other actions. AI headsets from mobile phone manufacturers often differ from mobile phone salesBinding, Huawei’s new AI headphones have sold more than 20,000 pairs in a single store on JD.com.

The second group is audio brands. These brands that originally make hardware such as headphones and speakers are subdivided into two categories. One category is those that have access to external companies such as Tencent, Baidu, and OpenAI, such as Philips and Newman. Large models or self-developed models (Sena Galaxy GPT4.0 large model) can be connected to the corresponding APP through headphones to realize translation, recording and translation and other actions. This type of headphones has been strongly associated with sports or meeting scenarios, has strong functionality, and has a high sales volume. Cleer ARC 3 headphones have sold more than 20,000 pairs in a single store on JD.com.

The other type, such as WISHEE and iKKO, are making a fuss about the headphone compartment, adding a touch screen and communication module, or building in ChatGPT to become an independent hardware device, but the current use experience is average. The sales volume of this type of headphones is relatively low, with sales in a single JD.com store ranging from a few hundred to a few thousand pairs.

Real photos of products from the official accounts of iKKO and WISHEE Xiaohongshu

The last category is the Internet giants and technology companies. These companies themselves have translation technology, voice technology and With the accumulation of large-model technology, the purpose of making AI headsets is to commercialize its own advantages through hardware in a closed loop.

For example, iFlytek, which participated in CES this year, specializes in meeting scenarios. It provides services such as recording, transcribing, translating, and generating meeting minutes through a voice recorder. Now it connects these services with mobile phones through AI headsets; another example is The Ola Friend headphones released after Byte acquired Oladance (headphone brand) are connected to Byte's Bean Bag intelligent assistant, extending the function of Bean Bag to daily companionship scenarios.

It is worth noting that for commercialization reasons, unlike Xiaodu and Yuanbao, which are open to other headphone brands, Byte has chosen a closed loop in software and hardware, stipulating that bean bags can only be used by Ola Friend headsets. Wake up, other headphones cannot wake up. Currently, more than 10,000 pairs of this headset have been sold in a single store on JD.com.

To summarize, compared with traditional wireless headsets, in addition to basic functions such as listening to audio and making and receiving calls, the standard direction of this batch of AI headsets is to serve as productivity tools, in meetings and business, etc. Translation and voice transcription functions are expanded in specific scenarios, but the degree of usability differs. Some headsets can also be connected to voice assistants, but the experience is stiff. Ola friend is almost the only AI headset that focuses on voice question and answer.

However, we can also implement these functions on our mobile phones, so why do we still need to use AI headphones?

Miyagi, a practitioner in the headphone industry, told "Fix Focus One" that the difference mainly lies in the subtle differences in user experience. First, the smart assistant can be woken up even when the screen is off, saving many steps; second, when it is inconvenient to operate the phone with both hands, there is no need to pause the thing at hand, for example, you can choose a restaurant while riding a bicycle; third, it involves many When in a scene, one person can wear a pair of headphones to achieve real-time translation.

It’s just that these headphones cost more than 1,000 yuan.More than 1,000 yuan, and some even exceed 2,000 yuan. When the AI ​​experience of many products is not as good as imagined, and some AI headsets emphasize functions at the expense of sound quality, whether consumers are willing to pay for them may vary from person to person.

The headphone industry has developed to this day and has become an industry that is so involution that it cannot have any shortcomings. For price-sensitive users, white-brand headphones priced at NT$9.9 are everywhere on major e-commerce platforms; for users who are not price-sensitive, headphones with good sound quality and design are also continuing to be iterated. AI has become almost the only upgrade point in the headphone industry.

But for now, AI-related functions are just the icing on the cake. Miyagi said that headphones must be easy to use first before talking about AI. AI is not yet a reason why consumers must order headphones.

AI headset sales are booming, how many hurdles are there?

As early as October 2011, since Apple first exposed its digital assistant Siri, the industry has reached a consensus on the development of voice interaction. But why, more than thirteen years later, has voice interaction still not become mainstream? Now with AI headsets, can Siri become a reality? At present, it seems that there is still some distance to go.

The biggest resistance is that the current AI headsets are "not smart enough" and "rely on mobile phones", which limits users' desire to purchase.

Wang Song, future smart CTO of iFlytek, once said that a real smart headset must be a device that can be independently networked and calculated independently, and can be used even without a mobile phone.

Miyagi explained that AI headsets are small in size, light in weight, and have limited computing and storage capabilities. They usually need to rely on large models such as mobile phones and other external devices to access the cloud to complete AI functions. Therefore, a common situation is that AI functions cannot be completed without a mobile phone. A more common situation is that even if a large model is connected, the AI ​​headset is not smart enough.

There are several key sticking points. For example, if you ask the smart assistant a question on your mobile phone or web page, the generated answer will be a long paragraph. If you listen to such an answer through headphones, you will not be able to finish it; For example, in your daily communication with other people, you often encounter interruptions or the other party's behavior of picking up the conversation. Therefore, if the AI ​​headset experience is to be more intelligent, headset manufacturers need to optimize large-scale model training, dynamic speech recognition, and voice interaction training.

At the same time, one thing that cannot be ignored is that users have not developed the habit of chatting with headphones.

Image source/Screenshot of iFlytek official website

Currently, the default social etiquette for wireless headphones is that wearing headphones means "do not disturb" or "resting". Most of the time In the office or on a business trip, you want a quieter, noise-isolated environment.

Many AI headsets want to create new scenes of "AI voice" companionship, but Miyagi pointed out that on the one hand, users are not used to talking to themselves after wearing headphones, and this will also be a ""Embarrassing me" scene, and I am not used to asking AI for everything; on the other hand, in order to achieve the effect of "accompanying", many AI headphones are made into open styles, so that users can wear them for a long time without ear canal discomfort, but This form has no sound insulation or noise reduction, and the use experience in noisy environments such as shopping malls is very poor, and it is difficult to balance noise reduction and experience.

However, Qingsong, an investor in the consumer electronics track, pointed out that optimistically, AI headphones can cultivate users’ awareness of daily use of AI better than AI glasses. “In the past, when using Kimi or Doubao, you had to open a web page first. Or APP, and then type and chat, you can now chat directly with the AI ​​through the earphones, which improves efficiency. As long as the AI ​​is easy to use and the answers are useful enough, the habit of chatting with the AI ​​at any time will gradually develop. ”

In addition, defects in mobile phone system permissions have become a “flaw” for third-party AI headset brands.

Qingsong said that when using third-party AI headsets, they must Prevent the mobile phone system from "killing" the headset APP running in the background due to operations such as power optimization. This will harm the user experience. Many people are also worried that almost all mobile phone manufacturers are releasing AI headsets, headsets and their own mobile phone systems. Will cooperate It is smoother, thus forming a strong binding relationship. How can third-party headset products compete with it?

Honor once ordered 2,000 cups of Luckin coffee using the smart assistant in the mobile phone at the press conference. This is The advantage of mobile phone manufacturers is that they can mobilize major applications in the system and go deep into the transaction process.

“If this can be realized through AI headsets in the future, it is actually what Alibaba wanted to do through Tmall Genie smart speakers. Tell stories. "Miyagi said, but he believes that this still goes back to the old problem of user habits. In real life, users have not yet formed the habit of waking up the assistant through headphones to help order coffee.

Who has the future in AI headsets? Who is the most embarrassed?

From a larger perspective, the rise of each generation of Internet giants is related to the competition for entrance. ——From search to WeChat and now AI, AI serves as a portal There is no doubt that it will become a battleground for military strategists.

Other hardware such as AI glasses is still a relatively new demand. It not only needs to cultivate people’s habit of wearing glasses, but also solves the problems of battery life and AI function optimization. question, but the novelty of the market and the possibility of becoming a standalone device, It has given many opportunities to entrepreneurs upstream and downstream of the industry chain.

As for headphones, almost everyone has headphones now. Adding AI functions to the already very mature headphone market is more likely to increase sales. Means, the possibility and necessity of headphones becoming independent hardware are also lower. p>

Qingsong explained that in terms of product design, although there are already independent headphone hardware products such as WISHEE and iKKO on the market, the small size and small screen of the headphone box can only realize limited functions and provide little incremental experience. Significantly increase power consumption; consumers’ mentality and usage habits,Headphone hardware that is completely divorced from consumer cognition has higher education costs.

At the same time, many investors told "Fix Focus One" that independent hardware may be a false proposition. Today, when the mobile phone experience is becoming more and more optimized, do we need another independent hardware besides the mobile phone? Think about it again, when AI Agent is really realized, will we still need hardware?

From the perspective of investors, there will be no disruptive innovation in the AI ​​headset industry. This also means that there are almost no small and medium-sized entrepreneurial opportunities in this track. The industry share will move closer to both ends of the price band, or it will be in vain. Brand, or big name.

Picture source/Screenshot of Huawei Mall official website

From this perspective, Miyagi analyzed that in the process of AI-based headphones, mobile phone manufacturers have the most advantage because they have Your own operating system has different levels of control capabilities. The most embarrassing thing is audio brands, especially traditional headphone brands. Their headphone products were not cheap because of their high sound quality and design. However, in recent years, they have been slow to transform in terms of AI functions and the prices are very high. Tasteless.

At this stage, rather than being optimistic about the AI ​​hardware industry, we are optimistic about the prospects of the mobile phone AI ecosystem and the penetration of AI in daily life.

Compared with third-party large model manufacturers or headset manufacturers, the advantage of mobile phone manufacturers lies in their ecological integration capabilities. Miyagi said that the integration here not only refers to the ability to cut into life service links, but also refers to the ability to link headphones with tablets, watches, smart home and other devices of the same brand, and to open up device permissions and user data to create a smoother experience. .

But mobile phone manufacturers also have their own problems. First, mobile phone manufacturers still need to improve the experience of large-scale model Q&A. Apple has currently adopted the method of accessing OpenAI (overseas markets) and Wen Xinyiyan (tentatively for the domestic market). Will domestic mobile phone manufacturers that do not have self-developed large models will Taking this approach remains to be proven. Second, each application in the mobile phone system needs to be balanced to what extent it will open up permissions to mobile phone manufacturers.

However, telling AI stories does bring a certain amount of imagination to the future of the headphone industry.

Qingsong said that one of the great imaginations is the possibility of charging on the software side. For example, there are limits on the number of free uses of functions such as listening to music, translation, and transcription, and some AI functions may directly become subscription functions in the future. In addition, as the capabilities of major domestic models improve, from search to chat to companion agents, the application scenarios of AI headsets will also expand and gradually penetrate into all aspects of users' lives.

In 2025, more AI-based consumer electronics will begin to accelerate. As Kai-Fu Lee said before: Today’s AI no longer requires Nobel-level innovation, but the productization and commercialization of existing technologies to create real value.

The same logic applies to the AI ​​headset industry.