Project Aria is Meta Reality Labs Research’s egocentric multimodal sensing platform, comprising research glasses worn like ordinary spectacles plus a desktop tools and machine-perception services suite. The Gen 1 device, distributed since 2020, includes an RGB camera, two SLAM cameras, eye-tracking cameras, an IMU, magnetometer, barometer, GNSS, and spatial microphones, with onboard recording of 1-2 hours per charge. Aria Gen 2, announced February 2025 and shipping to researchers in 2026, adds an upgraded sensor suite with on-device SLAM, eye tracking, hand tracking, and audio interactions plus expanded onboard compute. Over 200 academic and corporate partners use Aria glasses to collect ‘human data’ — first-person video and motion of people performing everyday tasks — which is increasingly used to bootstrap training for humanoid robots that need to learn the same tasks. A notable example is Georgia Tech’s EgoMimic, which used Aria recordings to achieve a reported 400% improvement in robot task success with just 90 minutes of data. Project Aria Tools is open-sourced, and the program operates under a research-access programme rather than a commercial product release, with strict privacy guidelines for wearer and bystander protection.
Meta Reality Labs' wearable research glasses with RGB and SLAM cameras, eye tracking, IMUs, GNSS, and spatial microphones. Used by 200+ academic and corporate partners to collect first-person 'human data' for training embodied AI. Aria Gen 2 ships to researchers in 2026.
Project Aria is Meta Reality Labs Research’s egocentric multimodal sensing platform, comprising research glasses worn like ordinary spectacles plus a desktop tools and machine-perception services suite. The Gen 1 device, distributed since 2020, includes an RGB camera, two SLAM cameras, eye-tracking cameras, an IMU, magnetometer, barometer, GNSS, and spatial microphones, with onboard recording of 1-2 hours per charge. Aria Gen 2, announced February 2025 and shipping to researchers in 2026, adds an upgraded sensor suite with on-device SLAM, eye tracking, hand tracking, and audio interactions plus expanded onboard compute. Over 200 academic and corporate partners use Aria glasses to collect ‘human data’ — first-person video and motion of people performing everyday tasks — which is increasingly used to bootstrap training for humanoid robots that need to learn the same tasks. A notable example is Georgia Tech’s EgoMimic, which used Aria recordings to achieve a reported 400% improvement in robot task success with just 90 minutes of data. Project Aria Tools is open-sourced, and the program operates under a research-access programme rather than a commercial product release, with strict privacy guidelines for wearer and bystander protection.