How much editing do you do to make your Narrations sound better?

My rule of thumb is do as little as possible to get the desired result for a particular recording, which begs the question,"What is the desired result"? Editing out unwanted sounds is mandatory, and for that I use two different techniques. I delete sounds a human being would never make (e.g, dogs barking, trucks going by). Breathing is normal, but perhaps distracting. I rapidly fade breath sounds down and then up by applying automation envelopes. Replacing breaths with silence can sometimes sound unnatural. It depends on how the track will be used. If I want more low-end, I use a mic that has a distinct proximity effect and move closer to the mic. If I want the most natural sound possible, I use a large-diaphragm condenser mic that has a flat frequency response. Thinking about these things in advance means I don't have to apply EQ and "fix it in the mix" during post-production. The final aspect of the desired result is the overall level. My target is between -6 and 0 db, preferably closer to 0. That's what normalization is for. Alternatively, I might use a mastering plug-in to adjust the input level as needed and limit the output level to my target. If the script isn't a "Sunday, Sunday, Sunday!" over the top announcer thing, there's not much to be gained (no pun intended) by applying compression.

Under most circumstances, I do not recommend noise gates for removing low-level unwanted sounds. Gating can sound unnatural and introduce its own artifacts. Summary: Get the best sound you can in the original recording and manipulate it as little as possible, bearing in mind the end use.