Skip to content
RobothecaBeta

GR00T Whole-Body Control (GEAR-SONIC)

  • AI model
  • Humanoids and robot learning

A humanoid whole-body controller framework from NVIDIA, with SONIC motion-tracking policies for Unitree G1 robots, plus teleoperation, planning, and training code.

  • Custom weights terms
  • Closed core

Made by NVIDIA Research, United States (checked on github.com/NVlabs/GR00T-WholeBodyControl/blob/main/LICENSE (opens in a new tab)).

Source on GitHub: github.com/NVlabs/GR00T-WholeBodyControl (opens in a new tab). Facts read at commit 1fa5f7f (opens in a new tab).

Licenses

Code
Apache-2.0 (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) at commit 1fa5f7f (opens in a new tab), on 30 Oct 2026. A license text Robotheca's reader does not know: a person at Robotheca set it on 5 Nov 2026.

Weights
NVIDIA Open Model LicenseTerms of their own, not a standard license

Read on huggingface.co/nvidia/GEAR-SONIC (opens in a new tab).

Commercial use
Unknown

Running the controller on a robot needs NVIDIA TensorRT, which is closed and under NVIDIA's own terms; the weights are under the NVIDIA Open Model License.

Parts it needs are closed

  • Needs NVIDIA Isaac Sim components (training): Training goes through Isaac Lab, which needs Isaac Sim's closed components.
  • Needs NVIDIA TensorRT (deployment): Running the controller on a robot needs NVIDIA's closed TensorRT at an exact version (10.7 or 10.13), under NVIDIA's own terms.

Labels

  • Custom weights termsThe weights come under their maker's own terms (NVIDIA Open Model License), not a standard license: read their conditions before any use.
  • Closed coreIt needs closed parts: NVIDIA Isaac Sim components (training) and NVIDIA TensorRT (deployment).

What it does: Robotheca's summary

  • SONIC is a motion-tracking foundation model that learns motor skills from large human motion data, aiming for one policy covering many behaviors.
  • Provides three released G1 checkpoints differing in reference lookahead and intended use: default, low-latency teleoperation, and SONIC v1.1.
  • Includes a C++ deployment stack that runs the policies on hardware, with TensorRT deployment and PyTorch checkpoints for Isaac Lab.
  • Supports VR whole-body teleoperation via a PICO headset and IsaacTeleop 1.4, and a kinematic planner for locomotion styles.
  • Offers SONIC training code for training from scratch or finetuning on the Bones-SEED motion capture dataset.

Written by Robotheca from the project's README (opens in a new tab) at commit 45c03fa.

Getting started

Difficulty
Advanced

Maintenance

Status
Active: a commit in the last 90 days
Newest commit
27 Oct 2026 on main
Latest release
None published on GitHub

Checked on GitHub on 30 Oct 2026.

Links

Robots that can run it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.