NVIDIA Isaac Teleop

NVIDIA's open-source teleoperation and data-collection framework, with unified support for XR headsets, gloves, and other input devices.

Isaac Teleop is NVIDIA's open-source (Apache 2.0) teleoperation and demonstration-data-collection framework; its first version was released in November 2025, bundled with Isaac Lab starting at Isaac Lab 3.0 Beta, and it also supports Isaac Sim, ROS 2, and real robots. It addresses the problem that every headset, glove, and robot combination otherwise needs its own integration code and produces data in its own format: it provides unified access to XR headsets such as Apple Vision Pro, PICO, and Quest, plus data gloves, foot pedals, and body trackers, all with synchronized timestamps; it uses composable retargeting modules to map human motion onto different robots; and it records and replays using MCAP, which interoperates with the LeRobot format. It already supports XR-controlled grippers or dexterous hands, seated whole-body teleoperation (HOMIE), SONIC-based whole-body teleoperation, and embodiment-free first-person data collection. According to GitHub, the repository was renamed NVIDIA/IsaacCapture in 2026, though the documentation still refers to it as Isaac Teleop.

ExampleWearing an XR headset such as Apple Vision Pro inside Isaac Lab, an operator teleoperates a simulated robot arm through a demonstration using hand motion; the recorded MCAP data is then converted into the LeRobot format for training a policy.

Also called

Isaac Teleop, IsaacCapture

Related

Teleoperation · VR Teleoperation · Motion Retargeting · NVIDIA Isaac Lab · MCAP · HOMIE

Sources

NVIDIA Isaac Teleop（GitHub，现为 NVIDIA/IsaacCapture） (Chinese)
Isaac Teleop Architecture 文档 (Chinese)

As of

2026-09