Two-in-One: Unified Multi-Person Interactive Motion Generation by Latent Diffusion Transformer

TL;DR

This paper introduces a unified latent diffusion transformer model for multi-person interactive motion generation, effectively capturing interactions and handling diverse motion differences guided by natural language, with improved quality and efficiency.

Contribution

It presents a novel single-model approach that models multi-person interactions in a unified latent space using VAE and diffusion, reducing complexity and enhancing generation quality.

Findings

Outperforms existing methods in generation quality.

Handles asymmetric motions effectively.

Speeds up the generation process.

Abstract

Multi-person interactive motion generation, a critical yet under-explored domain in computer character animation, poses significant challenges such as intricate modeling of inter-human interactions beyond individual motions and generating two motions with huge differences from one text condition. Current research often employs separate module branches for individual motions, leading to a loss of interaction information and increased computational demands. To address these challenges, we propose a novel, unified approach that models multi-person motions and their interactions within a single latent space. Our approach streamlines the process by treating interactive motions as an integrated data point, utilizing a Variational AutoEncoder (VAE) for compression into a unified latent space, and performing a diffusion process within this space, guided by the natural language conditions. Experimental results demonstrate our method's superiority over existing approaches in generation quality, performing text condition in particular when motions have significant asymmetry, and accelerating the generation efficiency while preserving high quality.