leroyjenkens said:
And the purpose of those is to basically get the sound closer to the person talking, correct?
Yes, that is correct. It is also true that you want to get the microphone closer to anything that you attempting to record, so long as its not so close that it causes distortion/clipping in the microphone itself (e.g., if you are recording an explosion, it's not necessary to put the microphone right next to the TNT. But for most other things, the closer the better).
When using an omnidirectional microphone, the recording sound intensity falls off at at \frac{1}{r^2}, where r is the distance between the camera and the whatever it is producing the sound.
For example, suppose the camera's internals make some sort of tiny hum, with 1 unit of power. Let's assume that the camera's microphone is 5 cm away from the source of the hum. Also suppose that there is a person 5 meters away that is the subject, and that person is talking with a sound power that is 10,000 units of power (10,000 times as loud as the camera's hum).
The recorded volume of the internal hum, in our arbitrary units, is \frac{1}{0.05^2} = 400.
The recorded volume of the person, in our arbitrary, units is \frac{10,000}{5^2} = 400
Which means the camera's internal hum sounds as loud as the person. Obviously, getting the microphone off the camera and closer to the person is advantageous.
There are other types of microphones besides omnidirectional, such as cardiod or hypercardiod (your on-camera microphone is probably a cardiod, I'm guessing), that have a little different characteristics in terms of directionality, but even with that, I'm sure you get the idea.
Like if the camera is on one side of the room near one person, then that person sounds fine, but a person on the other side of the room won't sound as good and will sound far away and maybe echoey, which makes it sound amateur.
There's a couple of options here.
Typically in this scenario, the camera will only be focused on one person at a time. The scene is shot multiple times with a given actor (usually the one that is speaking) being the subject of the camera. Not only is the camera repositioned, but the microphone is too. It means that the actors need to repeat the same scene over (and over and over sometimes). Then everything is edited together later in post, so it seems like there is only one conversation with no repeats.
For reasons of continuity, it is also a good idea to record the background noise (if there is any) separately. For example if they are in a restaurant, record ambient restaurant noises (without the microphone being close to anybody) for at least the duration of the scene. Then mix that audio in with the rest as the final step in post. (Also, if you have control, have all the other extras -- the other customer and waitstaff in the restaurant scene -- be dead quiet when you are filming the characters' conversation. Have the extras pretend to talk to each other, but while actually remaining silent. Only have them actually converse when you're filming the background track. Of course you might not be able to have this much control, but I'm just saying this is how it's actually done in the industry.)
On the rare occasion where both characters are to be talking in the same shot, place an omnidirectional microphone somewhere in-between them, or perhaps better, use a voice-over for one of the characters (only record one character at a time, and mix the audio together in post).
