Meta has pulled back the curtain on Aria Gen 2, its latest experimental smart glasses designed as a research hub for augmented reality (AR), artificial intelligence (AI), and robotics. Unlike consumer-ready wearables, these glasses are purely focused on giving researchers a sandbox to explore cutting-edge human-computer interaction. According to Meta, Aria Gen 2 aims to “unlock new possibilities for human-computer interaction” by upgrading nearly every aspect of the original 2020 Project Aria, their first research-only headset. Weighing in at around 75 grams and sporting a suite of sensors that rival many high-end smartphones, the new glasses could pave the way for future AR devices that blend seamlessly into everyday life.

Project Aria debuted in 2020 as Meta’s initial stab at a research-focused AR headset. Researchers worldwide used it to collect egocentric vision data, build AI models for spatial understanding, and prototype robotics applications. However, the first-gen glasses were relatively bulky and lacked some advanced sensors now considered essential for fluid AR experiences. Fast forward to February 27, 2025, and Meta revealed Aria Gen 2, which builds directly on lessons learned in the intervening years. This iteration preserves the core research mission—providing an open platform for academic and industry scientists—while dramatically upgrading the hardware. As Meta describes it, the new glasses will “shape the next computing platform” through improved sensor fidelity and more intuitive human-AI interfaces.

One of the first things you’ll notice about Aria Gen 2 is that it looks more like a refined pair of sunglasses than a clunky prototype. Meta has added folding arms—a first for their research glasses—to make them more portable. Despite packing in four computer vision (CV) cameras, eye trackers, and a photoplethysmography (PPG) sensor, the glasses weigh just 75 grams, roughly the weight of a pair of traditional prescription frames. They also come in eight different sizes to accommodate diverse head shapes and ensure that sensors align properly with users’ eyes and facial features. The inclusion of folding arms suggests Meta is already thinking about user comfort and real-world use cases, rather than just laboratory environments.

Perhaps the most significant upgrade over Project Aria is Aria Gen 2’s eye-tracking system. The new glasses employ dedicated cameras for each eye, enabling per-eye gaze tracking, blink detection, and precise pupil-center estimation. Why does this matter? By understanding exactly where users are looking—and even how their pupils dilate—researchers can infer intent, attention, and cognitive load. This deeper insight into visual attention could lead to AR interfaces that adapt in real time to what you find interesting. Meta claims that, with this advanced eye-tracking, they can better model “visual attention and intentions,” which is the holy grail for seamless human-AI collaboration.

Along with eye tracking, Aria Gen 2 is equipped with four outward-facing CV cameras that support 3D hand and object tracking. These cameras work in concert to capture six degrees of freedom (6DOF) movement, letting the device localize itself in space and detect users’ hand gestures with high precision. Meta envisions researchers using this data for tasks like “dexterous robot hand manipulation,” where a robot mimic human hand movements in real time. Beyond robotics, the same CV suite can power object recognition, scene understanding, and even early-stage spatial computing prototypes that require sub-millimeter accuracy.

A standout feature of Aria Gen 2 is the photoplethysmography (PPG) sensor embedded in the nosepad. By measuring subtle blood-volume changes under the skin, this sensor estimates the wearer’s heart rate without needing a wrist strap or chest band. This opens up new avenues for studying how physiological signals correlate with cognitive load or emotional state during AR experiences. Complementing the PPG sensor is a contact microphone also housed in the nosepad. Unlike traditional microphones that pick up background noise, this contact mic isolates the wearer’s voice—ideal for clear verbal commands in loud environments. Meta says these biosensing features could be used to build more empathetic AI systems that respond to stress or fatigue in real time.

Aria Gen 2 also packs an ambient light sensor, which can differentiate between indoor, outdoor, and mixed lighting environments. This allows software to dynamically adjust display brightness (in future AR-enabled versions) or tweak camera exposure for better image capture. On the audio front, the glasses feature open-ear, force-canceling speakers that deliver spatial audio cues without blocking environmental sounds. Researchers at Carnegie Mellon, for example, are already using these spatial audio features to help blind and low-vision users navigate indoor spaces via directional sound prompts. With clear audio input from the contact mic and rich binaural output, Aria Gen 2 becomes a platform for developing accessible AR experiences as well.

Rather than streaming data to the cloud, Aria Gen 2 uses custom-built silicon to run machine perception algorithms locally. That means six degrees of freedom simultaneous localization and mapping (6DOF SLAM), eye tracking, hand tracking, and speech recognition all happen on-device. On-device processing slashes latency—imperative when you want instantaneous responses to gestures or gaze shifts. According to Meta, this custom chip achieves a balance between power efficiency and performance, enabling six to eight hours of continuous operation on a single charge. In an era where battery life often bottlenecks wearable innovation, this is a significant leap forward.

Meta is distributing Aria Gen 2 to universities, research labs, and select companies. Already, Georgia Tech researchers have used the first-gen Aria to train robotic arms for household chores; the new glasses promise even finer control by leveraging the upgraded sensors. BMW has tested the Aria platform for potential in-car AR experiences, such as overlaying maintenance instructions directly onto engine components. Meanwhile, Envision—a company specializing in assistive tech—is experimenting with Aria Gen 2 to develop AI-guided spatial audio navigation for blind and low-vision users, supplementing work that Carnegie Mellon started with the first version. These diverse applications underscore Aria Gen 2’s role as more than just a glitzy prototype; it’s a catalyst for next-gen robotics, accessibility tools, and contextual AI research.

While Aria Gen 2 is strictly for research, it ties directly into Meta’s consumer-facing ambitions. The company’s partnership with EssilorLuxottica has already yielded Ray-Ban Meta smart glasses, which integrate minimal sensors and audio to deliver notifications and take photos. Looking ahead, Meta plans to roll out Orion AR glasses—boasting holographic displays—and even more advanced “Hypernova” glasses with built-in high-resolution screens. Rumors also swirl about a potential Oakley partnership to push ruggedized AR wearables. Data and insights gleaned from Aria Gen 2 will likely inform these consumer products, helping Meta balance performance, cost, and form factor as it edges closer to mainstream AR adoption.

Meta isn’t alone in the race. Apple is reportedly prioritizing a premium AR headset, slated for release around 2026, that could undercut Meta by leveraging its tightly controlled ecosystem. Google, on the other hand, has teamed with Samsung to develop Android XR glasses aimed at broad compatibility across devices. The battle won’t be about raw specs alone; challenges like weight, battery longevity, design aesthetics, and software ecosystems will shape user adoption. Whereas smartphones converged on similar slab-like designs, smart glasses demand fresh approaches to fashion, ergonomics, and seamless AI integration. With the EU’s Digital Markets Act potentially forcing Apple to open its platform, competition may hinge on innovation rather than walled gardens. If all goes as planned, we might finally see a vibrant marketplace where style, user experience, and developer ecosystems define success—rather than proprietary lock-in.

Meta has confirmed that Aria Gen 2 will be available to approved researchers later this year, with an application portal opening in mid-2025. No pricing details have been announced, but since these are intended as experimental research tools, they won’t hit retail channels. Instead, researchers can sign up on Meta’s website to request units, receiving early access to documentation and support. Looking further out, it’s reasonable to expect that insights from Aria Gen 2’s sensor suite—particularly the eye tracking and biosensing data—will feed directly into future consumer AR wearables. Whether that leads to AR glasses that can gauge your stress and adjust content delivery, or emergent robotics systems that anticipate your needs before you speak, the foundation being laid today with Aria Gen 2 could reshape how we interact with digital worlds tomorrow.