Project Aria
| Project Aria | |
|---|---|
| Basic Info | |
| VR/AR | Augmented Reality |
| Type | Egocentric research glasses |
| Subtype | Sensor-only data recording and machine perception device |
| Creator | Meta (Reality Labs Research) |
| Developer | Meta (Reality Labs Research) |
| Manufacturer | Meta |
| Announcement Date | September 16, 2020 (Gen 1); February 27, 2025 (Gen 2) |
| Release Date | Research distribution only (not sold to the public) |
| Price | Not for sale (provided to qualified researchers) |
| Website | https://www.projectaria.com/ |
| Versions | Aria Gen 1, Aria Gen 2 |
| System | |
| Storage | |
| Display | |
| Display | None (no see-through display) |
| Image | |
| Optics | |
| Optics | None |
| Passthrough | N/A |
| Tracking | |
| Tracking | Visual-inertial odometry (6DoF), SLAM |
| Eye Tracking | Yes (two inward-facing cameras) |
| Hand Tracking | Yes (Gen 2, on-device) |
| Audio | |
| Audio | Open-ear speakers (Gen 2); spatial microphone array |
| Microphone | Multi-channel spatial microphone array; contact microphone (Gen 2) |
| Camera | RGB point-of-view camera plus monochrome scene cameras |
| Connectivity | |
| Battery Life | 1.5 hours continuous recording (Gen 1); 6 to 8 hours continuous use (Gen 2) |
| Device | |
| Weight | About 75 g |
| Sensors | RGB camera, monochrome scene cameras, eye-tracking cameras, IMUs, magnetometer, barometer, GNSS, ambient light sensor, PPG heart-rate sensor (Gen 2) |
Property "Creator" (as page type) with input value "Meta Platforms|Meta]] (Reality Labs Research)" contains invalid characters or is incomplete and therefore can cause unexpected results during a query or annotation process. Property "Developer" (as page type) with input value "Meta Platforms|Meta]] (Reality Labs Research)" contains invalid characters or is incomplete and therefore can cause unexpected results during a query or annotation process. Property "Manufacturer" (as page type) with input value "Meta Platforms|Meta" contains invalid characters or is incomplete and therefore can cause unexpected results during a query or annotation process. Property "Tracking" (as page type) with input value "Visual-inertial odometry]] (6DoF)" contains invalid characters or is incomplete and therefore can cause unexpected results during a query or annotation process.
Project Aria is a research program and wearable sensor device built by Meta through its Reality Labs Research division. The device takes the form of a pair of glasses fitted with cameras and other sensors, but it has no display and shows nothing to the wearer: its purpose is to record egocentric (first-person) data so that researchers can study the perception, mapping and contextual artificial intelligence problems that future augmented reality glasses will need to solve.[1][2]
Meta announced the program on September 16, 2020 under its former name Facebook, and at launch stated plainly that the glasses are "not a consumer product, nor are they a prototype" and would not be sold to the public.[3] The device records video and audio from the wearer's point of view along with eye-movement and location data, and the resulting datasets are used to train and evaluate algorithms for simultaneous localization and mapping, eye tracking, hand tracking and audio understanding.[3][4]
A first-generation device (Aria Gen 1) was distributed from 2020 to a small group of Meta employees and contractors and later to external academic and corporate partners. A second generation, Aria Gen 2, was announced on February 27, 2025, with broader distribution to qualified researchers targeted for the second quarter of 2026.[5][6] By 2025 Meta said the glasses were in use internally and by more than 200 academic and corporate partners.[7]
Background and purpose
Meta's Reality Labs Research group develops the hardware and software it expects all-day wearable AR glasses to require, including perception that runs on a small, low-power, head-worn device. Project Aria exists to gather the real-world, first-person data needed for that work before any consumer AR product is ready to ship. Because the device only senses and records, and does not render anything for the wearer, Meta describes it as a data-collection and streaming platform rather than a head-mounted display.[4][2]
The 2023 device paper, "Project Aria: A New Tool for Egocentric Multi-Modal AI Research," authored by Jakob Engel, Kiran Somasundaram, Michael Goesele and a large group of Reality Labs Research collaborators, frames the program around egocentric machine perception and contextual AI, with the stated goal of supporting "always available, context-aware and personalized AI applications" for future wearable devices.[4]
History
Facebook (renamed Meta in 2021) introduced Project Aria on September 16, 2020. At launch the glasses were worn by about 100 Facebook employees and contractors, primarily in the San Francisco Bay Area and Seattle, who used them to collect data in everyday environments.[3] Alongside the internal rollout, Meta announced a pilot with Carnegie Mellon University's Cognitive Assistance Laboratory to build 3D maps of indoor spaces such as museums and airports, work aimed in part at navigation aids for people with visual impairments.[1]
Over the following years Meta opened the program to outside researchers through the Aria Research Kit and released a series of public datasets recorded with the device. The 2023 device paper formalized the Gen 1 hardware and its data formats for the research community.[4]
On February 27, 2025 Meta announced Aria Gen 2, calling it the next generation of egocentric research glasses and describing upgrades to sensing, comfort, on-device computation and audio interactivity.[7][8] Meta later said it was closing Gen 1 applications and, on October 19, 2025, opened applications for Gen 2 with separate forms for academic and corporate researchers, targeting a broad rollout to qualified applicants in the second quarter of 2026.[6]
Aria Gen 1
Aria Gen 1 weighs about 75 grams and carries five cameras: one rolling-shutter RGB point-of-view camera, two global-shutter monochrome scene cameras used for SLAM, and two inward-facing monochrome eye-tracking cameras.[9] The RGB camera reaches a maximum resolution of 2880x2880 pixels with a 110-degree horizontal field of view through an f-theta fisheye lens, while the two scene cameras run at 640x480 pixels with a 150-degree horizontal field of view.[9]
Non-visual sensing on Gen 1 includes two inertial measurement units (running at 1000 Hz and 800 Hz), a seven-channel spatial microphone array sampled at 48 kHz, a magnetometer, a barometer, GPS, and Wi-Fi and Bluetooth beacon sensing. Meta lists battery life as about 1.5 hours of continuous recording plus roughly 30 hours of standby in its lowest-power profile.[9]
| Component | Detail |
|---|---|
| Cameras | 5 total: 1 RGB, 2 monochrome scene (SLAM), 2 eye tracking |
| RGB camera | Up to 2880x2880 px, 110-degree HFOV, rolling shutter, fisheye lens |
| Scene cameras | 640x480 px, 150-degree HFOV, global shutter |
| Eye-tracking cameras | 2 inward-facing, monochrome, global shutter |
| IMUs | 2 (1000 Hz and 800 Hz) |
| Microphones | 7-channel spatial array, 48 kHz |
| Other sensors | Magnetometer, barometer, GPS, Wi-Fi and Bluetooth beacons |
| Weight | About 75 g |
| Battery | About 1.5 h continuous recording; about 30 h standby (low-power profile) |
| Display | None |
Aria Gen 2
Aria Gen 2 keeps the roughly 75-gram weight of the first generation (Meta cites about 74 to 76 grams depending on size, across eight size options) and adds foldable arms for easier storage and transport.[7][10] Battery life rises to six to eight hours of continuous use.[7][8]
The sensor suite is expanded relative to Gen 1. Gen 2 uses four computer-vision cameras instead of two, with global shutters and a high dynamic range of 120 dB, up from 70 dB on Gen 1, and a stereo overlap of 80 degrees versus 35 degrees on the earlier device.[10] It retains an RGB camera, eye-tracking cameras, a spatial microphone array, IMUs, a magnetometer, a barometer and GNSS.[7] New additions include a calibrated ambient light sensor, a contact microphone embedded in the nosepad that helps separate the wearer's voice from that of bystanders, and a photoplethysmography (PPG) sensor, also in the nosepad, that allows estimation of the wearer's heart rate.[10][8]
A custom low-power co-processor designed by Meta runs machine-perception tasks on the device itself, including 6-degree-of-freedom visual-inertial odometry, eye tracking and articulated hand tracking, as well as on-device speech recognition.[7][10] Gen 2's eye-tracking output covers gaze per eye, the vergence point, blink detection, pupil center and diameter, and corneal center.[10] The glasses also add audio output through open-ear speakers, which lets researchers prototype systems that respond to the wearer in real time, something the sensor-only Gen 1 could not do.[7]
| Feature | Aria Gen 1 | Aria Gen 2 |
|---|---|---|
| Computer-vision cameras | 2 | 4 |
| Camera dynamic range | 70 dB | 120 dB |
| Stereo overlap | 35 degrees | 80 degrees |
| On-device perception | No (offline processing) | Yes (custom co-processor: VIO, eye and hand tracking, speech) |
| Heart-rate (PPG) sensor | No | Yes (in nosepad) |
| Contact microphone | No | Yes (in nosepad) |
| Speakers | No | Yes (open-ear) |
| Battery (continuous) | About 1.5 h | 6 to 8 h |
| Arms | Fixed | Foldable |
Role in VR and AR
Project Aria sits upstream of Meta's augmented reality hardware: it is the data-gathering and perception-testbed layer for the technologies that AR and mixed reality glasses depend on, including head and eye tracking, hand tracking, environment mapping and contextual AI. Because the glasses have no display and are never sold, they isolate the sensing and perception problem from the rendering and product problem, letting Meta and outside researchers study how a small head-worn device can localize itself and understand its surroundings without the cost or constraints of a full AR headset.[4][2]
The same sensing stack that Aria exercises, visual-inertial odometry, SLAM, eye tracking and hand tracking, is the stack that drives positional tracking and interaction on consumer VR and AR devices such as the Meta Quest 3 and Meta's Ray-Ban Meta smart glasses, so progress made on Aria data feeds the broader VR and AR pipeline.[7][5] Coverage by Road to VR framed Gen 2 as a way for Meta to guide a large external research community toward the perception problems future AR glasses must solve, without Meta having to coordinate directly with every team.[5]
Aria also distinguishes itself from a consumer smart glasses product by its purpose. Meta's Ray-Ban Meta glasses are a shipping product with a camera, speakers and a voice assistant for everyday use, whereas Aria is a research instrument that records a far richer multi-sensor stream for algorithm development and is restricted to approved researchers.[8][2]
Datasets and research kit
Researchers access the device through the Aria Research Kit, which Meta offers to selected partners through a rolling application process. The kit includes the glasses, a companion app for controlling them, the open-source Project Aria Tools (projectaria_tools) for visualizing and processing recordings, and the Aria Studio desktop application.[11] Meta also provides Machine Perception Services that compute derived data from recordings, such as 6-degree-of-freedom trajectories and 3D eye-gaze estimates.[11]
The program has released several public egocentric datasets recorded with Aria glasses, used widely in computer vision and robotics research:
| Dataset | Focus |
|---|---|
| Aria Everyday Activities (AEA) | Everyday first-person activity recordings |
| Aria Digital Twin (ADT) | Ground-truth scenes (200 sequences across two instrumented indoor environments, 398 object instances) |
| Aria Synthetic Environments (ASE) | Synthetic indoor environments for perception training |
| Ego-Exo4D | First- and third-person skilled activity, a collaboration between Meta FAIR, Project Aria and 15 university partners |
| HOT3D | 3D hand and object tracking with eye gaze (3.7M-plus images, 19 subjects, 33 objects) |
| Nymeria | Full-body human motion in the wild |
| MMCSG | Multi-modal conversations recorded with smart glasses |
Meta describes the Ego-Exo4D dataset, collected with first-generation Aria glasses, as a foundational resource across modern computer vision and robotics research.[7][12]
Privacy and responsible use
Because the device continuously records video, audio and location data in real environments, Meta published a set of responsible-innovation measures alongside the 2020 launch. The glasses carry a prominent recording indicator and a physical mute control that stops data collection when pressed. Captured data is encrypted, kept in a separate designated storage space, and held under a quarantine period (Meta cited three days at launch) before researchers can access it, and it is automatically processed to blur faces and vehicle license plates.[3] Meta stated that wearers were instructed not to record in private spaces such as restrooms or in sensitive contexts.[3]
Current status
As of mid-2026 Project Aria remains a research-only program. Aria Gen 1 applications have closed, and Aria Gen 2 applications are open to academic and corporate researchers, with broad distribution of the Gen 2 device to qualified applicants targeted for the second quarter of 2026.[6][5] Meta has not announced any consumer sale of either generation; the glasses are positioned as a tool for the perception and contextual-AI research that underpins future AR glasses.[7][2]
See also
- Reality Labs
- Ray-Ban Meta
- Meta Quest 3
- Smart glasses
- SLAM
- Visual-inertial odometry
- Eye Tracking
- Hand Tracking
References
- ↑ 1.0 1.1 "Announcing Project Aria: a research project on the future of wearable AR". September 16, 2020. https://tech.facebook.com/reality-labs/2020/9/announcing-project-aria-a-research-project-on-the-future-of-wearable-ar/.
- ↑ 2.0 2.1 2.2 2.3 2.4 "Aria Gen 2, from Meta - Project Aria". https://www.projectaria.com/.
- ↑ 3.0 3.1 3.2 3.3 3.4 "Privacy Matters: Project Aria". September 16, 2020. https://about.fb.com/news/2020/09/privacy-matters-project-aria/.
- ↑ 4.0 4.1 4.2 4.3 4.4 "Project Aria: A New Tool for Egocentric Multi-Modal AI Research". August 2023. https://arxiv.org/abs/2308.13561.
- ↑ 5.0 5.1 5.2 5.3 "Meta to Ship Project Aria Gen 2 to Researchers in 2026, Paving the Way for Future AR Glasses". February 27, 2025. https://roadtovr.com/meta-to-ship-project-aria-gen-2-to-researchers-in-2026-paving-the-way-for-future-ar-glasses/.
- ↑ 6.0 6.1 6.2 "Aria Gen 2 Applications Are Now Open". October 19, 2025. https://www.meta.com/blog/aria-gen-2-updates/.
- ↑ 7.00 7.01 7.02 7.03 7.04 7.05 7.06 7.07 7.08 7.09 "Introducing Aria Gen 2: Unlocking New Research in Machine Perception, Contextual AI, Robotics, and More". February 27, 2025. https://www.meta.com/blog/project-aria-gen-2-next-generation-egocentric-research-glasses-reality-labs-ai-robotics/.
- ↑ 8.0 8.1 8.2 8.3 "Meta's new AR glasses for research can measure heart rate". February 27, 2025. https://techcrunch.com/2025/02/27/metas-new-ar-glasses-for-research-can-measure-heart-rate/.
- ↑ 9.0 9.1 9.2 "Hardware Specifications - Aria Gen 1 Docs". https://facebookresearch.github.io/projectaria_tools/docs/tech_spec/hardware_spec.
- ↑ 10.0 10.1 10.2 10.3 10.4 "Inside Aria Gen 2: Explore the cutting-edge tech under the hood". 2025. https://ai.meta.com/blog/aria-gen-2-research-glasses-under-the-hood-reality-labs/.
- ↑ 11.0 11.1 "Project Aria Research Kit". https://www.projectaria.com/research-kit/.
- ↑ "Ego-Exo4D Dataset - Aria Gen 1 Docs". https://facebookresearch.github.io/projectaria_tools/docs/open_datasets/ego-exo4d.