We Tested Google’s AI Glasses Prototype – One Step From Mass Production

Smart wearables are poised to dominate the future AI‑hardware battlefield, with glasses being the most critical product. At this year’s Google I/O, TechCrunch obtained an unreleased AI glasses prototype and performed a short‑term hands‑on test of the visual‑plus‑audio version.

The device was first announced at last year’s I/O and is a joint Android XR effort by Google, Warby Parker, Gentle Monster and Samsung. The eventual production model will blend each partner’s design language and embed the Gemini large model, enabling lens‑display of weather, walking navigation, Uber order info, real‑time translation, and user‑custom AI components, while supporting both iOS and Android phones.

The prototype focuses on display technology and flight validation rather than aesthetics, so its size, fit and build differ from the planned production unit. The mass‑production version will add automatic wear‑detect start/stop, a feature the prototype lacks.

To wake Gemini, users long‑press the right side of the frame for two seconds and hear a startup tone before beginning a conversation. In the demo, Gemini activation also launches the camera; the production model will expose a permission toggle for camera use.

During the first test, Gemini played a singer’s song. The noisy venue made details hard to hear even at maximum volume, showing the prototype cannot replace high‑quality Bluetooth earbuds but is adequate for casual listening. Its advantage is the lack of ear‑insertion, offering a more natural experience than AirPods in transparent mode. Tapping the middle of the frame pauses playback.

The second test involved photography. A short press of the photo button captures an image that syncs to the paired phone and watch. The prototype does not support video; the production version will add long‑press video capture with thumbnail preview. Users can also issue voice commands such as “take a photo and turn the person into anime style,” which sends the image to the phone, then to Gemini and Nano Banana servers for AI editing before returning the result.

Because the I/O venue’s Wi‑Fi load was high, the entire round‑trip for the photo‑edit workflow took about 45 seconds.

When the display is enabled, a minimal home screen appears below the line of sight. The demo pre‑installed weather and I/O countdown widgets, and users can add quick‑launch shortcuts for common apps like Google Maps and Translate. The prototype only has a single‑eye display, though the Android XR platform supports single‑display, dual‑display, and audio‑only configurations.

Testers noted the image on the lens was slightly blurry, likely due to wearing prescription lenses (one for distance, one for near). Covering one eye cleared the view, but after a short period the right‑eye area caused noticeable eye fatigue, whose cause remains uncertain.

The most impressive feature was real‑time translation. Leveraging the phone’s Google Translate, the glasses recognized spoken Spanish, displayed English subtitles on the lens, and Gemini read the translation aloud. The audio‑only version also supports translation, though subtitles appear on the phone instead of the lens.

Navigation was the second standout function. Although testers could not leave the venue, they observed that a voice command like “nearest coffee shop” triggers Gemini to launch Google Maps on the phone, load the route, and then display step‑by‑step turn prompts on the lens. The map appears at ground level with a blue location dot; turning the head rotates the map, mirroring the phone experience while keeping the forward view clear.

Object‑recognition queries were also tried. Initially the prototype failed to identify a Monet replica because the camera did not auto‑activate; after manually enabling the camera, Gemini identified the painting after two attempts. Subsequent queries (e.g., identifying plants or answering recipe questions) succeeded quickly, offering hands‑free functionality similar to Google Lens.

Google indicated that later this year, during an expanded trust‑testing phase, more details about the display version will be released. For now, the company is launching the pure‑audio version to satisfy some user needs, implicitly confirming that the display model is not yet ready.

The audio‑only glasses can also access Gemini via the frame’s speaker, allowing music playback, photo capture, calls, and app invocation. For example, during cooking the glasses can recognize meat in a pot and report whether it is cooked through.

Overall, Google’s AI glasses prototype has demonstrated all core consumer‑grade smart‑glasses functionalities. The remaining work focuses on refining comfort, display clarity, battery life and industrial design—the final hurdles before achieving a mass‑market product.

Image source: Google

Image source: Andrej Sokolow/picture alliance / Getty Images

Image source: Google

Image source: Google

Image source: Google

Image source: Andrej Sokolow/picture alliance / Getty Images

Image source: Google