Streaming Multi-Track Timeline Control for 3D Human Motion Generation

Excerpt

arXiv:2610.03873v1 Announce Type: cross Abstract: Text-driven human motion generation has advanced substantially, yet most methods assume instructions are available before synthesis. Interactive applications require responding to new instructions while continuing ongoing actions, such as answering a phone while walking. Existing approaches address streaming generation or simultaneous composition without explicitly combining streaming instruction arrival with independently timed, overlapping acti