The generation part of AI video got solved fast and then everyone stopped there, at a clip you still have to caption, cut, export, download onto a phone and post to three apps by hand.
So we built the rest of the loop instead: the captions cut from the voice model's own timings, the publish dialog, the stats coming back six-hourly, the remix button, and the schedule that runs the whole thing while you do something else.
It is a small product with sharp edges named on the page rather than hidden. That is on purpose, and it is the part we intend to keep.