PB-MLD: Part-Based Motion Latent Diffusion Model

J. B. Liu

doi:10.1016/j.imavis.2026.106118

PB-MLD: Part-Based Motion Latent Diffusion Model

2026-07-11

0

PRE

AI

DOI：10.1016/j.imavis.2026.106118

Original

Original request for help

Save

Abstract

En 中文

• Part-based framework: PB-MLD decomposes the human body into upper and lower parts for text-to-motion generation. • Structured modeling: Part-specific VAEs and InLineAttention model local and global motion. • High-quality synthesis: Generates globally coherent, physically plausible motions aligned with text.

PB-MLD: Part-Based Motion Latent Diffusion Model

Abstract

Journal

Organization