Waveguide smart glasses have a long way to go, but the $599 Rokid Glasses are the first pair I’ve tested that are refined and reliable enough to use with regularity. While they have a bit of a learning curve and their features are limited, they can provide closed captions and translated subtitles for users who are hard of hearing, as well as describe surroundings and translate text for users with poor vision. That said, they’re still far from feeling like a truly easy-to-use and comprehensive example of the technology, and I don't wholeheartedly recommend any waveguide glasses quite yet.  

What Are Waveguide Smart Glasses?

Waveguide smart glasses use tiny projectors and specially etched lenses to display information directly in front of your eyes, and connect wirelessly to your phone to process information and enable various features, such as voice control, language translation, and navigation.

I’ve reviewed two other waveguide smart glasses to date, the Even Realities G1 ($599) and the Vuzix Z100 ($499.99). Their functions are similar to those of the Rokid Glasses, but they are less polished in terms of usability, and their performance was inconsistent in my tests. They also lack cameras for computer vision, visual translations, and taking photos and videos, which is a significant advantage for the Rokid Glasses.

Meta recently announced its own take on waveguide glasses, the $799 Ray-Ban Display, which goes a step further than the Rokid Glasses with a full-color display instead of monochrome green and a new Neural Band wearable controller. I plan to review the Ray-Ban Display as soon as I can get my hands on a pair, so stay tuned for more.

All of the above-mentioned models are a completely different beast from the more developed prism-based display smart glasses, such as the RayNeo Air 3s Pro ($299), XReal One Pro ($649), and Rokid's own Max 2 ($449). These smart glasses are bulkier, wired, and function more like USB-C monitors you can wear on your face than anything else. I often use them when I want to work or watch a video away from home, as they effectively create a private virtual big screen in front of my eyes. Their projection systems put multiple layers of material between your eyes and the real world in front of you, and even when nothing is being shown on the glasses, they can still obstruct your view. They work well if you're sitting in a coffee shop or on an airplane, but it's not safe or comfortable to wear them while walking around.

Waveguide systems, on the other hand, provide an in-lens display that won't block your view. That way, you can wear the glasses anywhere, getting messages, instructions, and answers on the display, whether you’re at your desk or on the street.

(Credit: Will Greenwald)

Design: They Look Like Regular Glasses

At a glance, there’s no obvious indication that the Rokid Glasses are smart. They have simple rectangular frames made of glossy black plastic, with a flat face and slightly thick temples. The temples are mounted on spring hinges and both feature a small silver-colored Rokid logo on the outside, tiny black speaker grilles on the top edge, and pinhole microphones on the top and bottom. Otherwise, all of the glasses’ controls and connections are located solely on the right temple. That includes a camera button on the top edge near the hinge, and a touch-sensitive control strip identified by a thin ridge on the outside surface. A magnetic connector for the charging cable is located on the right eartip.

Only two clues will reveal their nature to anyone looking at the wearer. The first is a camera lens located in the upper-left corner of the front of the frame, which blends in with the black plastic but becomes apparent upon close inspection. The second clue is the lenses. They’re flat and clear, but the rectangular waveguides etched into them are visible because they can catch ambient light and reflect it at different angles and with different colors from the rest of the lenses.

The glasses are as comfortable to wear as any non-smart specs with similar frames. They weigh just 1.73 ounces (49 grams), and the spring hinges help accommodate users with larger heads. I had no issue wearing them for long periods of time, and was only reminded of the display and smart features when actively engaged with them.

(Credit: Will Greenwald)

If you’re nearsighted, you’ll probably have to get prescription lens inserts to use with the Rokid Glasses. Like most smart glasses with displays, prescription frames are included, which you can take to your optometrist to be filled. They have thin rims with tabs that slide into notches on the glasses, and small magnets that hold them in place between the waveguide lenses and your eyes. You should bring both the glasses and the prescription lens frames with you to the optometrist so they can see how they fit, and be prepared to spend more than you typically would for vision-correcting inserts. The holders sit very close to the waveguide lenses, and the optometrists I spoke with said that even for my mild prescription, high-index lenses that cost about twice as much as regular ones would be necessary to fit.

You’ll need a USB cable to charge the glasses, as the included adapter features a very short cable with a magnetic connector on one end and a female USB-C port on the other. In other words, you’ll need a standard male-to-male cable to actually plug the glasses into a USB port. It’s an odd choice, because while it means the adapter is small enough to easily tuck into the case, you still need a separate cable to use it, so there’s no real benefit to saving that space.

Speaking of the case, the Rokid Glasses come standard with a basic triangular case. A $99 charging case (free for Kickstarter backers) is expected to eliminate the need for dealing with the adapter. My unit only had the standard case, so I can’t speak to how effective or bulky the charging case is.

Hardware: Seeing Green, Hearing Voices

The Rokid Glasses use a green binocular display that projects a 640-by-480 picture through the waveguides, centered in your view. It has a 30-degree field of view, matching the Vuzix Z100, whereas the Even Realities G1 has a 25-degree viewing angle. As with the Vuzix Z100, I found it easy to align my eyes with the Rokid Glasses to see the entire projected image, which has been a problem with other waveguide-based smart glasses I’ve used. 

(Credit: Will Greenwald)

Because it’s all green, it’s not suitable for displaying any type of photo or video, only text and simple graphics. The green display is sharp and easy to read (with lens inserts, in my case). Its maximum light output is 1,500 nits, which is bright enough to see both indoors and outdoors in shade or overcast conditions. However, it’s difficult to make out in bright, sunny daylight, and there are no tinting options or sunglasses accessories to help with that.

A 12MP camera built into the Rokid Glasses enables computer vision for the AI assistant, allowing you to take photos and videos with a short or long press of the camera button located on the top edge of the right temple. The picture quality is comparable with the Ray-Ban Meta Smart Glasses ($299) or a low-to-midrange smartphone. Images are colorful and reasonably sharp in bright environments, but get very fuzzy in low light. When you take a picture or shoot a video clip, a bracket appears on the glasses’ display to indicate the center of the camera’s view, a helpful framing assist that the Ray-Ban Meta don’t have. The Even G1 and Vuzix Z100, meanwhile, don’t have any cameras at all.

(Credit: Will Greenwald)

The Rokid Glasses promise up to eight hours of power on a charge with mixed use, or up to six hours when streaming music over Bluetooth. This is a fraction of the claimed battery life of the Even G1 (1.5 days) and the Vuzix Z100 (two days), but it’s about in line with how long Meta says the Ray-Ban Display can last.

Like all smart glasses with audio, the Rokid Glasses use tiny speakers aimed at your ears to provide sound. Because of the physical gap between those speakers and your ears (unlike with earphones), you can hear everything around you as if you weren’t wearing the glasses. Conversely, any noise around you can easily drown out what you’re trying to listen to through the glasses. There’s also no bass to speak of, but the speakers otherwise sound pleasant from the low-mid to high frequencies if you’re in a relatively quiet location. I didn’t mind listening to music while walking down modestly busy streets in my NYC neighborhood, but I probably wouldn’t be able to hear anything on a subway platform. This audio weakness is typical across the entire category, and it’s not surprising to see it here.

Setup and Software: Pair to Your Phone, Tap, and Talk

To set up and use the Rokid Glasses, you need the Hi Rokid app, available for both Android and iOS. It manages most of the glasses’ features, using both your phone’s power and cloud processing. Pairing the glasses to your phone is easy: when they're folded, simply press the camera button on the right temple three times and tap the plus icon in the app. After that, you’ll be able to control the glasses, change settings, and update their firmware through the app. The Bluetooth connection between the Rokid Glasses and my Google Pixel 8 phone was reliable, rarely dropping or experiencing difficulty with automatic reconnection.

(Credit: Rokid)

Interacting with the Rokid Glasses is very simple: tap the touch strip to bring up the display, swipe left or right to switch between functions, and double-tap to exit a menu or turn off the display. Tapping also plays and pauses any music you’re listening to, and answers calls. The camera button lets you take a photo with a short press and a video with a long press at any time.

You'll probably control the glasses most often by voice, simply saying, "Hi, Rokid." Like the Even G1, the Rokid Glasses are built around a voice-controlled AI assistant and a few specific core functions you can activate manually with a basic visual interface. Tap the touch strip to display basic information, including the time, weather, and remaining battery life. Swipe right to see a menu with four options: Voice Translation, Prompter, Lyrics, and Settings. We’ll get to those features in a bit, but first, let’s discuss the voice assistant.

AI Assistant: Choose Your Model

Rokid’s AI assistant can answer general questions and perform tasks through the glasses and your phone. It’s accessible by saying, "Hi Rokid,” or by pressing on the touch strip. You can instruct the assistant to make a call, take a picture, record a video, translate a language, or simply answer a question. Thanks to the integrated camera, the AI assistant can describe whatever you’re looking at and even translate text.

The assistant uses your choice of two third-party AI models, ChatGPT or Qwen. ChatGPT is the ubiquitous model managed by San Francisco-based OpenAI, while Qwen comes from the Chinese company Alibaba Cloud. Both models support English, and you can individually select which one you want to use for language cognition and visual processing. In testing, I found both models work about as well as each other.

I had to get used to saying “Hi” instead of “Hey” because I’ve been trained by Google Assistant and Gemini to use the latter salutation, but now the AI assistant pops up reliably when I call for it. After the wake words, it lags for a second or two before listening for my request, and gives a similar pause before it can hear my responses when it asks for clarification. Using Google's voice assistant on my smart speaker or TV also requires waiting a moment between saying “Hey Google” and giving a command, but Rokid feels just a touch more sluggish. 

Even if it’s a little slow, the AI assistant provides useful information on demand. Asking a question brings up a ChatGPT- or Qwen-generated answer on the display, which is then read by your choice of male- or female-presenting voice. Visual processing is just as functional, and the glasses usually seem to accurately identify whatever I’m looking at on request.

The use of open third-party AI models means the Rokid Glasses do not interact with any wider information ecosystems, such as Alexa, Gemini, or Meta AI. The glasses won’t integrate with your social media feeds or email, and you can’t ask them to control your smart home devices. Having your queries run through ChatGPT or Qwen still means data flows through their servers, but they’re linked only to the Rokid account you need to register to use the glasses, and that account doesn’t directly connect to any of your other online accounts. Everything you do with Alexa, Gemini, and Meta AI feeds Amazon, Google, and Meta’s servers, respectively, with your name and any connected email accounts, purchases, and social media activity also flowing into it. Rokid’s AI assistant largely relies on the glasses themselves and doesn’t seem to significantly impact your personal digital footprint.

Features: Translations, Teleprompting, and More

Voice translation is accessible through the AI assistant or the glasses’ interface, and you can choose from a few different processing models: Qwen AI translation, Microsoft translation, or Rokid AI translation. Qwen and Microsoft are cloud-based, while Rokid translation can work offline. Offline translation requires a significant amount of processing power and is compatible with Android phones that have at least a Snapdragon 8 Gen 2 or an iPhone 14 or later; it doesn’t work on my Google Pixel 8. The other two use cloud processing.

The Qwen and Rokid translation platforms can automatically detect the language being spoken, with Qwen supporting 10 (Cantonese, Chinese, English, French, German, Italian, Japanese, Korean, Russian, and Spanish) and Rokid supporting six. Microsoft translation supports an impressive 89 different languages, but won’t automatically detect them; you’ll need to manually choose both the input and output languages through the phone app. The app also describes Microsoft translation as “free for a limited time,” though no details about any paid subscription plan are currently available.

(Credit: Rokid)

The translation feature works fairly well, under the right circumstances. The glasses use forward-facing microphones to capture the speaker’s voice, which must be loud and clear enough for speech-to-text processing to function properly before translation can occur. The results I got under various testing situations were mixed. I had to stand relatively close to colleagues or ask them to speak more clearly before the glasses would pick up what they were saying, at which point English was reliably transcribed and Spanish was translated. Outside, the glasses also accurately recognized and translated a Spanish-language sermon a street preacher was reciting through a speaker system across the street. It managed the Korean dub of the web novel Solo Leveling and the Japanese dub of Uglymug, Epicfighter reasonably well when I watched them at my desk with the speakers built into my monitor, though it stumbled when translating the Chinese dub of Lord of Mysteries with the same setup, and the Japanese dialog in One Piece I watched on my TV from my couch wasn’t picked up by the glasses at all.

Consistent cadence and properly delineating sentences and statements vastly improve any output the glasses produce. The feature attempts to process each sentence separately, and you can even see the translation actively change as the sentence progresses (which happens often when translating between languages with very different sentence structures, like Japanese and English). I noticed this when testing the glasses with a video by Ben “Yahtzee” Croshaw, who famously speaks at a constant clip through every episode of his appropriately named Fully Ramblomatic and Zero Punctuation series. The Rokid Glasses displayed an icon indicating that they were listening to the video as it played, but didn’t actually show any text until I paused the video after about 30 seconds, which made an entire paragraph appear before my eyes. The paragraph, “translated” from English to English, was to its credit a very accurate transcript.

You can choose from three different text sizes for the translations/captions and adjust the output's size and location on the display. This is helpful for keeping the text readable while not completely detracting from your vision, depending on your preferences.

Any voice translation you run will be recorded on the Rokid app, so you can review the audio and read the translations the glasses output at any time. This is very helpful, especially if you aren’t entirely sure of what was said at the time.

Despite the cumbersome requirements to get the translations to work, the end results were better than what I saw on the Even G1. The Vuzix Z100’s Google-based translation feature is even more reliable, but it lacks a microphone, relying instead on your phone’s mic. Its user experience is very barebones. Neither of those glasses automatically detects languages or records translations for later review.

In addition to voice translation, the Rokid Glasses can also translate text through the camera. You need to ask the AI assistant for this, and it uses your chosen visual processing model to do so. This works surprisingly well, once you get used to how large and clear the text needs to be for the glasses to read it. This can require moving your face closer to a screen or a piece of paper than you ordinarily would to read it yourself. After determining the light level and distance the camera requires, I consistently achieved good results. Both Japanese manga and Chinese text on a Sichuan restaurant’s menu were translated into English for me, with verbal readouts and green text on the display.

(Credit: Rokid)

The Rokid Glasses’ remaining features are much simpler. Teleprompter lets you load text into the app and have it scroll in front of your eyes for presentations and speeches. The Even G1 has a similar feature, and it can recognize what you’re saying and scroll the text automatically as you speak. It’s basic and useful.

Lyrics theoretically show you the lyrics of music you’re listening to through the glasses. I couldn’t get it to work with Spotify or YouTube Music on my Pixel 8. Currently, this feature appears to be supported only by Tencent’s QQMusic service, which is very popular in China but less common in North America.

Voice recording and transcription are also available on the Rokid Glasses, although unlike Voice Translation, Teleprompter, and Lyrics, they don’t have a button on the display’s interface. You’ll need to ask the glasses to start recording, and they will. After that, like with voice translation, both an audio file and text transcription will be available through the app.

You won’t get any real-time navigation features on the Rokid Glasses. At best, when asking for directions to a destination, the glasses will describe a loose path in a text paragraph, followed by a recommendation to use your phone’s map app for more accurate instructions. It seems we’re still far away from my dream of a video game-like minimap in the corner of my eye, though, considering the limitations of the glasses’ display, that’s for the best. The Even G1 has a map mode that lacks the resolution to even show street labels, and was effectively useless in my testing.

Accessibility: The Biggest Benefit

While the Rokid Glasses’ features are somewhat limited, a few of them make the glasses potentially very useful for a specific subset of users: people with hearing difficulties. English-to-English voice translation can provide live captions of anyone you’re speaking to, and it’s easier to use and more reliable than similar features on the Even G1 and the Vuzix Z100. 

I’ve had several readers ask me about live-captioning smart glasses, and I’ve informed them about the previous waveguide models, with several caveats regarding their learning curves and consistency. With the Rokid Glasses, it still takes some time to learn how to achieve the best results for voice captioning, but the feature generally works well. The overall smart glasses experience is significantly more polished than that of the Even G1 and Vuzix Z100, with a smoother setup process and a more reliable wireless connection. Those aspects are vital to make the glasses accessible to any users who want live captioning but aren’t the most tech-minded.

(Credit: Rokid)

There are some waveguide-based smart glasses available that focus entirely on captioning and translation, such as Captify, XanderGlasses, and Hearview. I haven’t used them, with the exception of trying an early version of Captify at CES this year, but they seem to be much simpler and more direct ways to access those features. As a major trade-off, they’re much more limited than any of the multipurpose waveguide smart glasses I’ve reviewed: XanderGlasses only provides live captions, and Hearview only provides captions and translated subtitles. They have no cameras, no AI assistants, and basically no other functions to speak of. The other trade-off is that two of them are very expensive: Hearview glasses cost $1,499 or $599, with the requirement of a $10 monthly subscription, and XanderGlasses cost $4,999. Even if the Rokid Glasses have a higher learning curve, they are user-friendly enough that I can recommend them for use with captions.

The camera and computer vision also make the Rokid Glasses potentially helpful for users with certain vision problems. You can ask the glasses to tell you what you’re looking at or read a document out loud for you, and they will. Poor eyesight might render the display useless to some, in which case camera-equipped audio-only AI glasses, such as the Ray-Bans Meta, are better suited for the task. On the other hand, the Rokid Glasses’ bright green text might be easier to see and read than many objects in inconsistent lighting, so there could still be value there. And again, if you’re concerned about privacy, you might prefer a single device’s use of a third-party AI model to Meta seeing everything you’re looking at.