In my view, this is a pretty good test of whether an image model was natively multimodally cross-trained with video. (The one in your brain, of course, was.)

Posters: davidad. davidad

3 likes recorded by the source corpus.