Thanks for the info Alex!
I tried the ffdshow normalization option, and it is a definite improvement. However, the normalization 'rides the gain' a couple of times a second, removing the dynamics from the video - quiet parts get louder, and loud parts get quieter. My goal is to have a fixed adjustment per video - like the 'replay gain' does for audio files, so that the dynamics of the music in the video doesn't get squished by the constantly adjusting normalization.
Like I said, this is an improvement, and is certainly better than turning the option off when running in an un-manned mode where you don't want huge volume changes. So thanks! But if anyone has ideas on how to adjust the volume by a fixed amount on a per-video basis, without the dynamics-robbing tracking normalization turned on, I'd love to be able to do that.
Another thing I noticed. I'm playing back on stereo speakers. When playing AC3 encoded videos that have surround, the volume still goes way up. Looking at the FFDSHOW audio decoder in the MIXER tab, you can see that all of the channels are getting mixed to the stereo pair at a full 1x multiplier. Thus the front left is getting not only the full left front channel, but also the left rear, the center, and the LFE. The mixer has a "normalize matrix" checkbox. When I select that, it drops the multiplier to 0.25 (12 db lower volume). That makes the volume too low. Any ideas on how to get FFDSHOW to automatically encode any stereo or surround file down to a stereo downmix at approximately the same volume level?
Thanks!!!