The robot side of the stack · facts from the GR00T-WBC model card

SONIC: where MotionBricks motion meets a real G1

MotionBricks answers "what should this character do next?". GEAR-SONIC answers "make a 30 kg robot actually do it". Both ship inside the same repository, and the README is explicit that they anchor NVIDIA's GR00T Whole-Body Control plan together.

How the two projects relate

The README's Related Work section states it in one sentence: "Together with MotionBricks, GEAR-SONIC anchors NVIDIA's GR00T Whole-Body Control initiative." In practice that means one shared monorepo (NVlabs/GR00T-WholeBodyControl), MotionBricks as the motion-generation layer (how the latent models work), and SONIC as the tracking controller that turns reference motion into motor behaviour on the Unitree G1.

GEAR-SONIC teaser: the Unitree G1 executing whole-body motion tracked by the SONIC controller — Official GEAR-SONIC teaser GIF from the shared assets folder. Open original

Official SONIC sources: project page · nvidia/GEAR-SONIC on Hugging Face · parent repo.

The three checkpoints

Verbatim from the GR00T-WBC model card — every checkpoint is in the nvidia/GEAR-SONIC Hugging Face repo:

Checkpoint	Reference lookahead	Intended use	Download
Default SONIC (original release)	10 future frames at 20 ms spacing, approximately 200 ms of reference lookahead	Default general-purpose SONIC controller for motion tracking, planning, teleoperation, and compatibility with existing deployments. G1 and teleoperation future-reference observations use step5.	nvidia/GEAR-SONIC · main
Low-latency teleoperation	4 future frames at 20 ms spacing, approximately 80 ms of reference lookahead	Intended for more responsive whole-body teleoperation and VLA execution. G1 and teleoperation future-reference observations use step1. Use its encoder, decoder, and observation config together.	nvidia/GEAR-SONIC · low_latency
SONIC v1.1	10 future frames at 20 ms spacing, approximately 200 ms of reference lookahead	Uses robot-heading-normalized target orientation and was trained with wrist-pose augmentation. Intended for heading-stable whole-body teleoperation and SONIC-backed VLA policies that use this controller. G1 and teleoperation future-reference observations use step5; this is not the low-latency model.	nvidia/GEAR-SONIC · sonic_v1_1

Shared specs: All three models use the SONIC universal-token controller, produce 64-dimensional latent motion tokens, run the controller at 50 Hz, and support SMPL pose, G1 motion reference, and teleoperation inputs.

Caveat, in the model card's own words: The lookahead values describe the reference horizon presented to the controller. They are not measurements of total end-to-end teleoperation latency, which also includes sensing, networking, preprocessing, and inference.

SONIC in the release timeline

The parent repo's news feed, SONIC-relevant entries only (full feed on the news page):

2026-05-07 — GR00T-WBC: end-to-end VLA workflow on G1 (teleop data → Isaac-GR00T N1.7 fine-tune → SONIC deployment).
2026-06-16 — GR00T-WBC: low-latency teleoperation checkpoint (4-frame lookahead) + Isaac Teleop Setup docs.
2026-07-23 — GR00T-WBC: SONIC v1.1 checkpoint (robot-heading-normalized, wrist-pose augmentation).

MotionBricks' own 2026-04-27 preview release is the starting point of the combined story; the checkpoint rows above are what arrived on the SONIC side afterwards.

Where to read next

MotionBricks vs NVIDIA's stack

How the pieces of GR00T (and what is NOT part of it) sort out.

Architecture

The VQVAE / pose / root models on the generation side of the pair.

News timeline

Every dated entry from both READMEs, with sources.

SONIC FAQ

What is SONIC in relation to MotionBricks?

SONIC (released as GEAR-SONIC) is NVIDIA's whole-body-control controller for the Unitree G1, shipped in the same GR00T-WholeBodyControl repository as MotionBricks. The README's Related Work section states that GEAR-SONIC and MotionBricks together anchor NVIDIA's GR00T Whole-Body Control initiative: MotionBricks generates the motion, SONIC tracks and executes it on the robot.

Which SONIC checkpoint should I use?

The original Default SONIC for general motion tracking, planning, teleoperation and compatibility with existing deployments; the low-latency checkpoint (4-frame, ~80 ms reference lookahead) for responsive teleop and VLA execution; SONIC v1.1 (robot-heading-normalized target orientation, wrist-pose augmentation) for heading-stable teleoperation. All three live in the nvidia/GEAR-SONIC Hugging Face repository.

Does the 80 ms lookahead number mean 80 ms of total teleoperation latency?

No, and the model card says so explicitly: the lookahead values describe the reference horizon presented to the controller, not end-to-end latency, which also includes sensing, networking, preprocessing, and inference. Any post quoting SONIC 'latency' from the lookahead table is misreading it.