I would like to share my personal remedy for videos/movies with barely audible 'whisper dialogues' followed by deafening explosions. While there is no one-size-fits-all setting for compression, I am hoping that this at least points you in the right direction or that you find it practical.
audio filter, compand, attack 0, release 1, -90 db in/inf out (gate), -70 db in/out, -21 dB in/out, 0 dB in, -15 dB out, soft knee default 0.01, 12 dB makeup gain, input level 0 dB, lookahead 0. Nasty transients seem to 'escape' the compressor if you omit "-90/-900" (I don't know why).
If this functionality had a GUI, it would probably resemble Adobe Audition's dynamics processing (comes with a graph).
Learning audio compression is tricky, and I found myself learning it and then re-learning it a couple more times. I guess it would be a dream come true if ffmpeg supported VST plugins like tube-saturation/brickwall limiting (e.g. Ferric TDS).
For a convenient ffmpeg GUI, google: "en masse, batch processor for videos" and follow the yo*tub* link.