I tried the new SOTA text-to-video models, they're overhyped