Stochastic Video Generation with a Learned Prior

Emily Denton, Rob Fergus

Abstract: Generating video frames that accurately predict future world states is challenging. Existing approaches either fail to capture the full distribution of outcomes, or yield blurry generations, or bot...