EMOSH: Expressive Motion and Shape Disentanglement for Human Animation

2026-06-26

AI Summary

Existing controllable human animation methods struggle to simultaneously achieve high expressiveness and disentanglement between motion and body shape: 2D pose-driven approaches often leak source identity shape, while 3D prior-based methods fail to model fine-grained facial expressions and hand gestures. To address this, this work proposes EMOSH, a framework that introduces an explicitly disentangled Expressive Human Model (EHM) as the control representation and incorporates a coarse-to-fine hybrid motion injection strategy with a spatial alignment conditioning mechanism. EMOSH is the first method to enable high-fidelity control of expressions and gestures while preserving strict geometric disentanglement, effectively eliminating shape leakage and enhancing consistency between training and inference. Experiments demonstrate that EMOSH outperforms existing approaches in both self-driven and cross-driven scenarios, producing videos with strong identity consistency, vivid expressiveness, and rigorous motion–shape disentanglement.

Abstract

High-fidelity and expressive controllable human animation is essential for content creation and digital avatar applications. However, existing methods face a dilemma between expressiveness and disentanglement. Mainstream 2D pose-conditioned approaches suffer from "motion-shape entanglement", leading to the leakage of the driving subject's body shape. Conversely, methods relying on 3D priors (e.g., SMPL) achieve geometric disentanglement but struggle to capture facial expressions and complex gestures, resulting in rigid animations. To this end, we propose EMOSH, a novel framework for high-fidelity controllable human video generation. First, an Expressive Human Model (EHM) is introduced as the core control representation. By explicitly disentangling shape and pose parameters, we fundamentally resolve the body shape leakage issue. Alongside this, a robust motion tracker is designed to accurately estimate EHM parameters from video. Second, we propose a Coarse-to-Fine Hybrid Motion Injection strategy, enabling more fine-grained control over expressions and gestures. Furthermore, we introduce a Spatially-Aligned Conditioning mechanism to bridge the domain gap between training and inference, improving identity consistency. Extensive experiments demonstrate that EMOSH outperforms previous methods in both self-driven and cross-driven scenarios, producing high-fidelity videos with vivid expressions while maintaining shape disentanglement.

Problem

• motion-shape disentanglement
• expressive human animation
• body shape leakage
• controllable video generation
• facial expressions

Innovation

• disentanglement
• expressive animation
• human motion modeling
• shape-pose separation
• video generation