Thanks to David Harbsmeier for this very useful guide.
While there's no one group of settings that will work in every situation, the Compressor/Limiter filter should help with persons that start each sentence speaking strong/loud and end each sentence speaking soft/quiet. First set the gain level for the clip to the level that sounds natural for when they're speaking soft/quiet. Apply the Compressor/Limiter filter and set the Threshold to -3 or -6db lower than what you want your average level to be. For example, if you want an average output level of -12, try a Threshold of -15 or -18 to see which works best. The Ratio setting that I've found to work best is around 6 or 8, depending on the subject, of course. For speech, I usually leave the Attack and Release setting at their default. Play the clip and watch the meters while listening critically to the beginning of a sentence where the speaker was strong/loud. Play around with the Threshold and Ratio settings until you get the desired amount of compression on the loud parts without affecting the quiet parts at all. The goal, of course, is to keep the vocals sounding realistic without spiking the meters. As with all filters, a little adjustment can go a long, long way.Once you have found the settings that work, you can apply that filter w/settings to all other clips for that speaker and you should be good.The Dynamic Processor filter also includes compression settings and I found it very helpful when compressing loud vocals without losing intensity or 'perceived volume' (e.g.; car commercials where it is apparently required by law for the VO talent to shout). That filter is also my secret weapon when clients bring in footage that was recorded too low. The Master Gain setting is a lifesaver sometimes.-DH