EchoAvatar: Real-time Generative Avatar Animation from Audio Streams 事件

EchoAvatar: Real-time Generative Avatar Animation from Audio Streams arXiv:2605.28272v1 Announce Type: new Abstract: Real-time synthesis of high-fidelity 3D character motion from audio is a pivotal component for next-generation interactive avatars and virtual assistants. However, most existing approaches are limited to offline processing of complete audio sequences or are constrained to specific domains, rarely handling both speech and music effectively. In this paper, we introduce a novel fram