The idea behind replaygain is you decide on a reference level you want your library to conform to. When you apply replaygain to a file your software should analyze the average levels of the signal and then calculate a gain level to be applied to make that track (or album) average to the desired level. It then adds one or more tags to the file that simply defines the amount to adjust the level of the output so that the average output level is what you want.
heres what ive done to tracks that sound way too low , ive increase the decibels and converted the flac file to flac (click on file > convert to> replay gain > move the preamp together> convert), […] what i do is keep my replay gains between -6db to -9db
precisely the point of RG is to do the work of levelling them for you, and it does this scientifically rather than estimating as you have been
Players could have another loudness stage when user pin-points/presets each frequency and measure/apply all frequency points as equaly loud (this takes human + equipment nicely into one check), based on that replaygain/r128 can do its work.
This may weirdly fail in multi-user listening situations.
p.s. Another thing that is not really calculated is level of love for a certain song (i guess "better" songs could need a dB or two).
This is interesting! I wouldn’t be surprised to see something like this emerging eventually (somewhere).
Yeah, I imagine the combination of differing responses in hardware and in the user could quite easily confound RG
I do wonder, however, if my perception of loudness could be accurately predicted by an algorithm. Such a beast might have to combine the equal loudness curve with the dynamic range and try to measure attacks, if that's even possible.