doesn't your TV have a setting for this? dynamic compression? so that there's not such a difference between voices and loud audio?
the hacker's way would be to pull out the audio track, run it through a sound editor like SoundForge, normalize the audio, and then reinsert the audio track back into the video file. this way, you're not having to re-encode the video (which would reduce quality and take forever, depending on how long the video is).