Volume Normalization

[podonnell]

I think this is handled differently depending on the player, but I am having a widely awful time with audio normalization.

Right now I'm using the jellyfin media player on 'Track Gain', which should squash the dynamic range. The alternative was 'Album Gain', which maintains the dynamic range. I'm not sure how that one would work, but it sounds like it would be worse.
I'm having to blast my audio up to 70 / 100 for certain dialogues, and I nearly had a heart attack when a bullet was shot at one point.

I feel like every device I have, I need to keep the remote in my hands.

I am ideally going to upgrade one or two of my systems where I watch, and get a proper 5.1 setup so I can keep dialogue in the center channel, but some of my devices are going to remain just simple TVs or connected to 2 external speakers.
For these, I really need a better solution.

Are there any better ways to handle this? Does it depend much on the audio codec / track that comes with the files? Any other tips?

[TheDreadPirate]

My understanding is that normalization does not change dynamic range of a track just peak or average volume. And that normalization in Jellyfin is only done on Music. My rudimentary coding skills seems to confirm that when I look at the normalization job.

Movie audio is mastered way differently than music and has huge dynamic range comparatively. When Jellyfin transcodes audio, there is a setting in Dashboard > Playback to apply a gain to the transcoded and, usually, down mixed audio. I'm not aware of any settings in Jellyfin to compress the audio, but a lot of sound bars and TVs have settings to change dynamic range. "Night" or "TV" mode will compress the audio so that the dialogue and loud sounds are much closer in volume.

[podonnell]

My understanding is that normalization does not change dynamic range of a track just peak or average volume.  And that normalization in Jellyfin is only done on Music.  My rudimentary coding skills seems to confirm that when I look at the normalization job.

Movie audio is mastered way differently than music and has huge dynamic range comparatively.  When Jellyfin transcodes audio, there is a setting in Dashboard > Playback to apply a gain to the transcoded and, usually, down mixed audio.  I'm not aware of any settings in Jellyfin to compress the audio, but a lot of sound bars and TVs have settings to change dynamic range.  "Night" or "TV" mode will compress the audio so that the dialogue and loud sounds are much closer in volume.

You might be right on about this being intended for music. At least in the Jellyfin Media Player specifically, that section is listed right under Music Quality, and 'track / album' make a lot of sense when I think about music now:

Audio Normalization

Track Gain
Track gain - adjusts the volume of each track so they playback with the same loudness. 
Album gain - adjusts the volume of all the tracks in an album only, keeping the album's dynamic range.

One of the players I use (maybe HPV Shim or my Hisense Jellyfin app) have an 'experimental night mode', but that didn't do too much either.

I think I'll have to check into this more from the TV / device side to see what options I have.

[LordMelvin]

Sorry to be a couple months late, but for posterity, what you are describing is a textbook case of not mixing a film or show's surround mix down to stereo properly. Surround sound mixes usually put almost all dialogue in the center channel with very little dialogue, but plenty of music and sound effects placed in the left and right side channels, and the left and right side channels will be the default two chosen by a discarding stereo implementation, where the player is accepting all audio channels from the server but then only feeding 'stereo' to the actual speakers.

The fix for this should be, on your device, in 'user' 'settings', under 'playback' assigning 'Maximum Allowed Audio Channels' to stereo. This should then ensure that the server is properly asked to downmix to stereo, with whatever downmix settings you have in place there.