AnchorRoute: Human Motion Synthesis with Interval-Routed Sparse Contro
AnchorRoute is aimed at the annoying gap between fully authored animation and vague procedural motion: you give the system a few sparse anchors, and it fills in the rest of the body motion. The paper’s main twist is that the same anchors are used twice. During generation, they’re converted into condition features and fed into a frozen Transition Masked Diffusion prior via AnchorKV and dual-context conditioning, so the model keeps the quality of the pretrained text-to-motion system while learning sparse spatial control.
After generation, those same anchors become residual signals for refinement. Their timestamps define correction intervals, and RouteSolver projects updates onto piecewise-affine interval bases to push the motion closer to the requested path. The authors report better anchor adherence than prior sparse-control methods under the sparse keyjoint protocol, while preserving text-motion quality. It’s especially relevant for animation tooling and runtime authoring workflows...
“the generator preserves text-motion quality, while RouteSolver provides a controllable path toward stronger anchor adherence”
- what
- AnchorRoute is a sparse-anchor human motion synthesis framework that supports root-3D, planar-root, and body-point control.
- who
- Paper by Pengcheng Fang, Tengjiao Sun, Dongjie Fu, Xiaoyu Zhan, Yanwen Guo, Hansung Kim, and Xiaohao Cai.
- when
- Submitted on 14 May 2026; revised 15 May 2026 (v2).
- impact
- Could help animation and gameplay teams author controllable character motion from a few sparse inputs instead of dense keyframes.
Promising control improvement without sacrificing motion quality
Follow motion synthesis updates
See relevant stories in your personalized news feed.
Discussion