CreativeVision

新着順 人気順

Learning to Act with Visual Priors

Creative Vision · Research Notes Learning to Act with Visual Priors Exploring how visual priors and prediction can improve action prediction, with robotic manipulation as a downstream testbed. Core contributors: Chen Zhu1 and Gang Yan2 Pro…

Streaming Speech-Driven Video Generation for Real-Time Conversational Characters

(Japanese version follows English one/英語版の後に日本語版が続きます。) Contributor Core contributor: Chen ZhuProject Lead: Shiqi YangCreative Vision Team, Multimodal AI Department Introduction Video generation is cool and has reached a re…

Few-Step World Action Models: Action-Side Distillation of Fast-WAM

(日本語版は英語版の後に続きます。) Contributor Shiqi YangCreative Vision Team, Multimodal AI Department Motivation World Action Models (WAMs) have recently emerged as a promising direction for embodied intelligence [1, 2, 3]. They connec…