Seeing Sound, Hearing Images: The Temperature of AI Glasses

With only half of 2026 gone, the number of AI glasses released has already reached 68, with brands such as HUAWEI, Thunderbird, XREAL, and KEDAXUNFEI all launching new products. From office work and sports to health assistance applications, AI glasses are showcasing their multi-functional capabilities.

It's worth noting that "audiovisual assistance," a niche feature of AI glasses, has been gaining attention from manufacturers in recent years. To some extent, it's not about the market's potential size, but rather the unique value that AI glasses can demonstrate within this market.

In January of this year, the Romanian startup .lumen showcased a disruptive accessibility technology—the .lumen navigation glasses for the blind. In February, Meta advanced AI glasses assistive functions to help people with cognitive impairments and veterans regain independent living; in March, Daemyung Optics of South Korea signed an agreement with Everysight to jointly develop AI glasses to provide solutions for people with hearing and visual impairments; in June, Rokid stated that it had established a dedicated team for special groups to conduct needs research among relevant user groups, and its products have been included in the list of assistive tools subsidized by the Disabled Persons' Federation in Zhejiang, Sichuan, Yunnan and other regions of China.

Making sound visible: Subtitle glasses have been included in China's "Assistive Device Subsidy" list.

For hearing-impaired users, the most direct benefit of AI glasses at present is the conversion of sound into text that can be read in real time.

This direction is not new. In 2022, LLVision released the AR subtitle glasses "Listener," which uses a microphone to collect conversational audio and then displays the transcribed subtitles in the user's field of vision. This product was also one of the earliest consumer-grade AR subtitle glasses to achieve mass production at the time.

With the maturation of speech recognition and lightweight display technologies, real-time captioning is beginning to move from specialized assistive devices into general-purpose AI glasses. Meta Ray-Ban Display supports displaying dialogue captions and translations within the lenses, with the text visible only to the wearer. Products like Rokid Glasses and Thunderbird X3 Pro have also incorporated real-time captioning and translation as key features of their display-based AI glasses.

Another technological approach is to enhance the sound itself. Meta's "Conversation Focus" feature, tested on Ray-Ban Meta and Oakley Meta HSTN, uses directional microphones to highlight the speaker's voice in front of the user, reducing interference from ambient noise. This type of solution is closer to mild hearing aids and is suitable for noisy environments such as restaurants, exhibitions, and train stations, but it cannot replace caption display for users with severe hearing impairments.

From a product perspective, AI glasses already possess the basic requirements for converting speech to subtitles, but there are still significant differences between accessible subtitles and ordinary meeting transcription. Everyday conversations often involve multiple people speaking simultaneously, dialects, accents, background music, and sudden environmental noise. In addition to recognizing text, subtitle glasses also need to determine the speaker's identity and voice direction, handle sentence breaks, tone, and context, and involve microphone data collection. Developing this function is easy, but doing it well is not.

Currently, Chinese national policy has begun to embrace this product form. The new version of Beijing's catalogue for subsidies for the purchase of assistive devices for the disabled has added "subtitle glasses" to the list. In July 2026, during an AI-powered accessible glasses experience event organized by the Haidian District Association for the Deaf in Beijing, real-time subtitles, intelligent translation, and navigation were the main features showcased.

By making the visuals audible, AI glasses are beginning to take on the role of environmental understanding.

The needs of visually impaired individuals for AI glasses are more complex. Low-vision users may retain some vision, and they require features such as image magnification, edge enhancement, contrast adjustment, color enhancement, and text recognition. For totally blind or nearly totally blind users, the display screen has relatively limited value; products need to convert image and spatial information into voice, tactile, or other cues.

This also determines two different product directions for assistive glasses for the visually impaired. The first type helps users "understand what's in front of them." Cameras can recognize text, goods, banknotes, people, bus and road signs, and then broadcast the results through open speakers. Multimodal large models improve scene description capabilities, allowing glasses to go beyond recognizing single objects and answer questions such as "What's on the table?", "What is the person in front of me doing?", and "What does this notice say?".

The second type helps users "determine their next move." These products need to identify road boundaries, steps, obstacles, and dynamic vehicles, determine the spatial relationships between objects, and then provide action suggestions through voice, vibration, or pressure feedback. .lumen adopts this latter approach. Its system borrows from the environmental perception concepts of autonomous driving, using sensors and AI to find safe walking paths and then using haptic feedback to guide users to change direction. The device does not rely on pre-drawn environmental maps, nor does it need to continuously announce left and right turns via voice, reducing the burden of auditory prompts on the user's auditory attention.

However, the responsibilities of a product increase significantly as it evolves from "describing the environment" to "guiding travel." While general AI question answering may occasionally exhibit recognition errors, potentially resulting in an incorrect answer, inaccessible navigation systems that misjudge steps, vehicles, or obstacles could directly impact user safety. Transparent glass, low-lying obstacles, suspended objects, backlighting, and nighttime environments are all scenarios where computer vision systems are prone to misjudgments. Existing research generally considers obstacle detection, path planning, sensor fusion, and interactive feedback as the core capabilities of navigation systems for the visually impaired.

Therefore, at this stage, AI glasses are more suitable as a supplement to canes, guide dogs, and human assistance. During the 2026 London Marathon, some visually impaired runners used Meta smart glasses to obtain landmark recognition and voice information, but the main guidance was still provided by accompanying runners or guide dogs. Real-world use cases show that AI glasses can improve information acquisition efficiency, but they are still far from fully supporting safe navigation.

Accessibility products also need to retain channels for human assistance. When the device cannot determine complex road conditions, users should be able to quickly contact family members or volunteers to transmit first-person perspective images and location information to remote assistants. AI handling high-frequency, standardized information, with human intervention in low-frequency but high-risk scenarios, may be a more reliable product combination at this stage.

Battery life: A major hurdle for accessible AI glasses

Accessible AI glasses are high-frequency assistive devices, and compared to ordinary consumer AI glasses, they have higher requirements for battery life and wearing comfort. This is because these products are not occasional assistive tools, but rather devices used frequently in high-frequency scenarios such as travel and communication. Therefore, battery life and weight are important indicators for the long-term use of accessible AI glasses.

Displays, cameras, and continuous AI computing remain the biggest power consumers. Meta Ray-Ban Display officially states a maximum battery life of 6 hours with mixed use; AI shooting glasses that partially eliminate the display screen can extend battery life to over 10 hours, but continuous video recording time can only be maintained within 1 hour, making it difficult to meet the needs of continuous AI visual understanding.

Currently, manufacturers are continuously exploring improvements in battery life. For example, Rokid's neckband battery in its Rokid Glass3 B2B business provides at least two and a half hours of continuous operation while the camera is working.

Furthermore, a replaceable battery design is another effective way to increase battery life from a hardware perspective. Users can remove and insert the temple to replace the main battery, and the secondary battery can maintain continuous power during the swapping process. When used with the battery swapping case, it enables all-day battery life. At the system level, a dual-chip architecture design can also significantly improve battery life.

Of course, besides the solutions listed above, a true revolution in the underlying materials of batteries is needed to genuinely solve the problem, such as advancements in silicon-carbon anode materials and solid-state batteries.

Related data shows that compared to ordinary graphite anodes, silicon-carbon anodes can significantly improve the mass energy density and volumetric energy density of batteries, typically increasing overall battery energy density by 8% to 40%. Simultaneously, due to silicon's high theoretical specific capacity, battery life and fast-charging performance can also be greatly optimized.

This material was first used in high-end smartphones and other electronic devices, and later it was also applied to glasses. The 263mAh XIAOMI Jinshajiang battery built into XIAOMI's AI glasses uses the same high-density silicon-carbon anode technology as that in XIAOMI phones, which greatly improves the battery life of the AI glasses.

Compared to traditional lithium-ion batteries, solid-state batteries offer higher energy density, faster charging speeds, and better safety. For AI glasses, the miniaturization and high energy density of solid-state batteries are particularly important. As wearable devices, AI glasses have very limited internal space. Solid-state batteries can better adapt to these compact space requirements while providing sufficient energy to support various functions of AI glasses, such as high-brightness displays, image recognition, and long-duration video recording.

Finally, the World Health Organization's 2019 World Report on Vision shows that more than 2.2 billion people worldwide suffer from visual impairment or blindness. Regarding hearing, approximately 1.5 billion people globally have varying degrees of hearing loss, with about 430 million experiencing severe hearing impairment.

While the absence of barriers may not be the fastest-growing market for AI glasses, it is certainly one of the scenarios that best tests the product's value. Ordinary users can tolerate occasional recognition errors and readily use their phones as alternatives; however, for users who rely on devices for communication and travel, accuracy, latency, battery life, and reliability directly impact user experience and personal safety.

In terms of accessibility, in addition to people with visual and hearing impairments, the "silver-haired" population has also attracted the attention of Chinese manufacturers. At last year's Senior Expo, LIANHUI Technology launched the world's first visual aid glasses equipped with the Om intelligent brain, designed specifically for the elderly. The glasses weigh only 45 grams, have an 8-hour battery life, and support AI capabilities such as sensing everything, proactive interaction, and autonomous decision-making.

At the 4th China Shenzhen International Optical Fair and the 2026 First Artificial Intelligence Eyewear Exhibition, MINGZHI Cloud Technology also launched AI-powered glasses for the elderly. Equipped with large-scale model technology, these glasses integrate functions such as vision assistance, hearing assistance, emergency response, and remote assistance, providing a customized smart wearable solution for senior citizens.

The human touch in AI glasses is constantly increasing.

Source:：VRTUOLUO