Sound made with the picture

A model that generates its own audio makes the sound and the picture together, so a door closing lands on the frame where the door closes. A model that takes an audio reference works to a track you supply, which is a different job and a different result.

Both are useful and they are not interchangeable. If you want ambience and effects that belong to the shot, you want the first. If you have a voice-over and the mouth has to match it, you want the second.

On a model that makes its own audio, naming the sound is not an afterthought, it is half the shot. "The latch clicks and the room goes quiet" describes something the model can actually produce.

Leave it out and it will invent something anyway. What it invents is usually generic, because that is what is left when nothing was asked for.

Several models price audio separately: the same second of video costs more with sound on than with sound off, sometimes twice as much. On those models the switch is a price decision, not just a quality one.

The studio works the amount out from the settings you actually chose and puts it on the button. If you turn audio off, the number moves before you press anything.

Silent is a legitimate choice. A lot of video is watched without sound, and a clip that was never given audio is cheaper than one that was and then muted.