Experience is a vision mix. The picture cuts between single shots of each host and a wide two shot of both. A fixed centre crop keeps the middle third of the width, which looks right on a single shot and lands in the gap between the hosts on a two shot.
Correct length. Correct aspect ratio. Captions burned in. Queued against the right channels. Every automated check passes, because nothing that reads metadata can see the picture.
So the pipeline runs a face detector across the finished file and counts faces frame by frame, and a contact sheet goes in front of a person before anything uploads.
That is one of a set of gates built over six episodes of running this show properly. Each one exists for a reason, each one is written down, and we will walk you through the lot in a first conversation. It is a more useful document than any feature list.