Stochastic Video Generation with a Learned Prior

Remi Denton, Rob Fergus

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 ge…