Topology aware transformer edits motion for various humanoid skeletons
TAME:Topology-Aware Text-Driven Motion Editing across Heterogeneous Humanoid Skeletons
Computer Vision and Pattern Recognition
Summary
Editing motion using text commands is hard when characters have different skeleton shapes. The authors created TAME, a new method that understands each character’s bone structure and changes motion accordingly. Their system uses smart attention techniques to focus on relevant joints and time frames, improving the quality of edits. They also built a new dataset to test motions on both familiar and new skeleton types, showing their method works better than others.
What this means in practice
- •For animation studios: Modify character motions with text commands across different skeleton designs without redoing animations from scratch.$Commercial implications: Enables software products that simplify animation editing for games and movies, supporting multiple character rigs.
- •For game developers: Adapt pre-made humanoid motions to diverse character skeletons in games quickly using natural language instructions.
Authors
Qichen Zheng, Siyuan Yang, Chong Wang, Jun Liu, Shijian Lu, Alex Kot, Kwok-Yan Lam
Abstract
Text-driven motion editing modifies an existing motion sequence according to a text instruction while preserving the content of the source motion. Existing methods are typically built for a single, fixed skeletal topology, which limits their use in animation pipelines where characters differ in joint count and skeletal hierarchy. We present Topology-Aware Motion Editor (TAME), a flow-matching transformer that edits motions on humanoid skeletons of varying topology. TAME represents motion as per-joint, per-frame tokens and models interactions among joints, across frames, and with the text instruction through skeletal, temporal, and text cross-attention layers. To make the skeletal attention follow each character's hierarchy, TAME replaces full joint attention with Topology-Constrained Skeletal Propagation (TCSP), which restricts attention to one-hop kinematic neighbors in the skeleton's adjacency matrix. We further introduce Edit-Focused Representation Alignment (EFRA), a self-distilled representation alignment strategy that aligns student features with cleaner EMA-teacher features exclusively on edit-relevant joint-time tokens, making edits faithful to the instruction. To make this setting trainable and comparable, we construct TopoMotionFix, a multi-topology extension of MotionFix with seen- and unseen-topology evaluation protocols. TAME outperforms previous methods in edit alignment and source preservation on MotionFix and reliably edits motions on unseen skeletons in TopoMotionFix.