Style machines

TLDR

Motion capture sequences exhibit distinct choreographies and styles. The study aims to learn motion patterns from diverse motion capture data to enable stylistic motion synthesis. The authors learn common choreographic elements and style variations, identifying a few stylistic degrees of freedom that capture dataset diversity. The resulting model can synthesize novel motion across style interpolations and extrapolations, convert novice ballet to expert modern dance, and be driven by video, scripts, or noise to generate new choreography.

Abstract

We approach the problem of stylistic motion synthesis by learning motion patterns from a highly varied set of motion capture sequences. Each sequence may have a distinct choreography, performed in a distinct sytle. Learning identifies common choreographic elements across sequences, the different styles in which each element is performed, and a small number of stylistic degrees of freedom which span the many variations in the dataset. The learned model can synthesize novel motion data in any interpolation or extrapolation of styles. For example, it can convert novice ballet motions into the more graceful modern dance of an expert. The model can also be driven by video, by scripts or even by noise to generate new choreography and synthesize virtual motion-capture in many styles.