2 accepted papers
We have recently seen tremendous progress in realistic text-to-motion generation. Yet the existing methods often fail or produce implausible motions with unseen text inputs which limits the applications. In this paper we present OMG a novel framework which enables compelling motion generation from z…