# GRAIL: Generating Humanoid Loco-Manipulation from 3D Assets and Video Priors

arXiv: https://arxiv.org/abs/2606.05160 · HTML: https://mlsanity.com/paper/2606.05160/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 3 Jun 2026.

## Built on by (most cited first, 6 of 6)

- [R2S-EGO: Dual-Proxy Refinement for Sparse-Capture Real-to-Sim](https://mlsanity.com/paper/2608.06827/) (2026)
- [RoboTok: A Scalable Data Engine for Internet Demonstration Video Retrieval and Dexterous Manipulation Learning](https://mlsanity.com/paper/2609.03199/) (2026)
- [Weave: Learning Whole-Body Dexterous Loco-Manipulation from Human-Object Interactions](https://mlsanity.com/paper/2609.16683/) (2026)
- [InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation](https://mlsanity.com/paper/2610.02196/) (2026)
- [COSMI: COmpositional Synthesis of Multi-object Interactions](https://mlsanity.com/paper/2610.03252/) (2026)
- [InterMimicGen: Scaling Humanoid Loco-Manipulation through Self-Evolving Motion Imitation](https://mlsanity.com/paper/2610.06850/) (2026)

## Abstract (arXiv)

Scaling humanoid loco-manipulation requires robot-compatible demonstrations across diverse objects, whole-body motions, and scene geometries, but teleoperation and motion capture are difficult to scale because each collection depends on physical setups, instrumented actors, and robot operation. We present GRAIL, a digital generation pipeline that remains fully virtual until deployment: it composes 3D assets, simulator-ready scenes, and priors from video foundation models (VFMs) to synthesize interactions without rebuilding physical environments or teleoperating the robot. Rather than reconstructing unconstrained in-the-wild videos, GRAIL starts from fully specified 3D configurations in which object geometry, camera parameters, metric scale, environment depth, and a robot-proportioned character are known before video generation and reused during reconstruction. This privileged setup better conditions 4D recovery, allowing model-based object tracking, human motion estimation, and interaction-aware optimization to reconstruct metric 4D human-object interaction (HOI) trajectories with reduced depth ambiguity and morphology mismatch. We retarget the recovered motions to a humanoid robot and train complementary task-general trackers: an object-aware latent adaptor for manipulation and a scene-aware tracker for terrain traversal. GRAIL produces over 20,000 sequences spanning pick-up, object manipulation, sitting, and terrain traversal. Using only GRAIL-generated data, we train egocentric visual policies through a sim-to-real pipeline and deploy them on a Unitree G1 humanoid, achieving 84% real-world success on diverse object pick-up and 90% success on stair-climbing.