Getting video models to learn better, faster