NVIDIA open-sources MotionBricks: real-time motion in-betweening at 1.3-2.5ms per request
maier_ak · x · 2026-11-20
MotionBricks — NVIDIA's SIGGRAPH 2026 model that fills in motion between poses in real time, demonstrated on game characters and the Unitree G1 humanoid — is now something you can run: pretrained human and G1 models, a Python runtime, demos, and authoring editors are released.
- Speed: one request takes 1.3-2.5ms on an RTX 4090 and yields 0.6-1.9s of motion
- License: code is Apache 2.0; models allow commercial use under NVIDIA's open model agreement
- Limits (per the author): no text or scene input (apps place targets and handle obstacle avoidance), output is kinematic and not validated for real robot control, and no quality benchmark of the released models yet — a fast motion generator, not a controller
Related event: NVIDIA Open-Sources MotionBricks, a Real-Time Motion Inbetweening Model(6 posts)→
More from Embodied
- NEO-GOTCHI A.I. hands-on: transparent-shell AI Tamagotchi with voice chat and AR — Megalania34 · 2026-11-20
- Inhabitant: training-free LLM agent completes 74/75 long-horizon household tasks — Yao Zhang · 2026-11-20
- Scoble lists 10 ways Tesla FSD fails at airport pickups ahead of wheel-less Cybercab sales — Scobleizer · 2026-11-20
- EPFL team unveils unified 3D generative model for structure-based drug design and docking — HannesStaerk · 2026-11-20
- GXO's 1,000-robot Apptronik deal is a letter of intent, hinging on 4 years of KPIs — YvesMulkers · 2026-11-20
- NVIDIA's SOMA Retargeter adapts human motion to five humanoid robots, from G1 to AgiBot — jonstephens85 · 2026-11-20