Learning to Generate Diverse Dance Motions with Transformer

TL;DR

This paper presents a novel transformer-based system for generating diverse, high-quality dance motions from music, leveraging a large YouTube-based dataset and new evaluation metrics, suitable for virtual concerts and professional animation.

Contribution

Introduces a two-stream motion transformer model trained on a large YouTube dance dataset, enabling flexible and diverse dance motion synthesis from music.

Findings

Outperforms state-of-the-art dance motion generation methods

Demonstrates effectiveness of online videos for training dance models

Provides high-quality dance animations for virtual events

Abstract

With the ongoing pandemic, virtual concerts and live events using digitized performances of musicians are getting traction on massive multiplayer online worlds. However, well choreographed dance movements are extremely complex to animate and would involve an expensive and tedious production process. In addition to the use of complex motion capture systems, it typically requires a collaborative effort between animators, dancers, and choreographers. We introduce a complete system for dance motion synthesis, which can generate complex and highly diverse dance sequences given an input music sequence. As motion capture data is limited for the range of dance motions and styles, we introduce a massive dance motion data set that is created from YouTube videos. We also present a novel two-stream motion transformer generative model, which can generate motion sequences with high flexibility. We also introduce new evaluation metrics for the quality of synthesized dance motions, and demonstrate that our system can outperform state-of-the-art methods. Our system provides high-quality animations suitable for large crowds for virtual concerts and can also be used as reference for professional animation pipelines. Most importantly, we show that vast online videos can be effective in training dance motion models.