Why isn't sound volume specified via two parameters: target and range?
In this way, when watching a movie or listening to a song, you can avoid uncomfortably loud sounds while still being able to hear soft sounds.
Hope this is a problem ML might solve some day with audio quality feedback. I place a little device with microphone next to me and it tells my player/speakers whether the music/sfx are too loud and the voices can't be heard, or a song in a playlist requires a different equalization/volume to sound right, or I just moved away and need the sound louder, then moved near and corrects it so I don't get deaf.
Some movies have been playing with having a wider dynamic audio range, and some are just mumbling. I don't think many listeners care for more than an audio compressor can give them and the "advertising industry" doesn't like that so it tends to not show up as a mass market thing.
If I understand you correctly this is called "Night Mode" or something similar on many TVs, phones and other audio devices