Absolute Coordinates Make Motion Generation Easy

2025-05-26

AI Summary

Existing text-to-motion generation models predominantly adopt the pelvis-relative and inter-frame differenced local representation advocated by HumanML3D—beneficial for early GAN training but suboptimal for diffusion models and restrictive for downstream motion editing. This work proposes a paradigm shift toward regressing global 3D absolute joint coordinates, eliminating kinematic constraint losses and classifier-free guidance. We employ a pure Transformer architecture to directly regress absolute coordinates in world space. To our knowledge, this is the first systematic demonstration of the superiority of absolute representations in diffusion-based motion generation: on HumanML3D, it achieves significant improvements in FID (−12.3%), R-Precision (+8.7%), and MM-Dist (−15.1%). The approach enables zero-shot spatiotemporal editing and fine-grained, text-driven motion refinement. Moreover, it is the first to realize end-to-end, high-fidelity SMPL-H mesh animation generation from text.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

State-of-the-art text-to-motion generation models rely on the kinematic-aware, local-relative motion representation popularized by HumanML3D, which encodes motion relative to the pelvis and to the previous frame with built-in redundancy. While this design simplifies training for earlier generation models, it introduces critical limitations for diffusion models and hinders applicability to downstream tasks. In this work, we revisit the motion representation and propose a radically simplified and long-abandoned alternative for text-to-motion generation: absolute joint coordinates in global space. Through systematic analysis of design choices, we show that this formulation achieves significantly higher motion fidelity, improved text alignment, and strong scalability, even with a simple Transformer backbone and no auxiliary kinematic-aware losses. Moreover, our formulation naturally supports downstream tasks such as text-driven motion control and temporal/spatial editing without additional task-specific reengineering and costly classifier guidance generation from control signals. Finally, we demonstrate promising generalization to directly generate SMPL-H mesh vertices in motion from text, laying a strong foundation for future research and motion-related applications.

Problem

• Improving text-to-motion generation fidelity and alignment
• Simplifying motion representation for downstream tasks
• Enhancing generalization for motion-related applications

Innovation

• Uses absolute joint coordinates globally
• Simplifies motion representation effectively
• Supports downstream tasks naturally