Motion style transfers adapt to different body shapes without losing detail
MorphoStyle: Motion Style Transfer with Morphology Control
Computer Vision and Pattern Recognition
Summary
People move in ways influenced by their actions, how they do them (style), and their body shapes. Previous methods could change how a movement looks (style) but only worked well for average body types, or they changed movements to fit different shapes without keeping the style consistent. The authors created MorphoStyle to solve this by separating style, action, and body shape in how it learns motion. This lets it change how a motion looks while properly fitting movements to bodies of different shapes, keeping the action and style distinct and plausible.
What this means in practice
- •For game developers: Generate character animations that adapt stylistic motions to characters of varied body shapes without losing movement realism.$Commercial implications: This enables game companies to produce diverse avatars with consistent stylish animations, enhancing player experience and visual diversity.
- •For animation studios: Create stylized motions for animated characters with different body types while preserving intended style and natural movement.
Authors
Xin Feng, Eleonora D'Arnese, Mohan Sridharan
Abstract
Human motion may be viewed as a combination of action content, style, and body morphology. Existing motion style transfer methods transfer a reference style onto a content motion while assuming a canonical body, whereas shape-aware motion generators adapt motion to a target shape without explicit style control. This separation of motion style and shape (morphology) makes it difficult to generate stylized motions for non-canonical bodies; naively combining a style transfer module with a shape-aware generator often leaks action content from the style reference and disrupts shape-consistent kinematics. We present MorphoStyle, a framework for shape-aware motion style transfer that is built on a shape-conditioned Finite-Scalar-Quantization Variational Auto-Encoder (FSQ-VAE). The key contribution is to pose the desired style transfer as modular latent disentanglement comprising: (i) a contrastive style encoder that extracts content-decoupled style embeddings; (ii) a text-guided style-routing mechanism that locates style-relevant joints in a text-motion feature space; and (iii) a manifold preserving style modulator that injects discriminative style embeddings in content features as a temporally-gated low-rank offset. Extensive experiments on benchmark datasets demonstrate that MorphoStyle outperforms competing baselines in terms of both shape control and motion style transfer, while simultaneously providing quantitative shape control. For more details, please see project website: https://github.com/funkdub/MorphoStyle.