Style-ERD: Responsive and Coherent Online Motion Style Transfer

TL;DR

This paper introduces Style-ERD, an online motion style transfer model that enables real-time, high-quality stylization of character motions with minimal latency, outperforming previous offline methods in realism and efficiency.

Contribution

The paper presents a novel Encoder-Recurrent-Decoder model with a combined attention discriminator for online motion style transfer, capable of handling multiple styles with improved realism and runtime performance.

Findings

Outperforms previous offline methods in motion realism.

Provides significant gains in runtime efficiency.

Supports multiple target styles with a unified model.

Abstract

Motion style transfer is a common method for enriching character animation. Motion style transfer algorithms are often designed for offline settings where motions are processed in segments. However, for online animation applications, such as realtime avatar animation from motion capture, motions need to be processed as a stream with minimal latency. In this work, we realize a flexible, high-quality motion style transfer method for this setting. We propose a novel style transfer model, Style-ERD, to stylize motions in an online manner with an Encoder-Recurrent-Decoder structure, along with a novel discriminator that combines feature attention and temporal attention. Our method stylizes motions into multiple target styles with a unified model. Although our method targets online settings, it outperforms previous offline methods in motion realism and style expressiveness and provides significant gains in runtime efficiency