Motion-Omni generates real-time full-body co-speech motion aligned with spoken dialogue

PKU1898 · hf · 2026-09-07

Motion-Omni is an end-to-end framework that jointly generates spoken dialogue and full-body co-speech motion from shared hidden states. It uses scalable pseudo-labeling to overcome data scarcity and a unified evaluation protocol, targeting real-time, aligned responses for embodied and digital-human applications.

Original post →

More from Embodied

Embodied channel →