Consider switching to ReplayGain 2.0

[derat]

nup update currently uses the mp3gain program to compute gain adjustments. It sounds like rsgain implements ReplayGain 2.0, which may be more accurate. I don't know if there are any speed differences or other concerns. rsgain looks like it isn't officially packaged for Debian (and I didn't find an ITP either), which makes me less excited about switching to it.

[derat]

loudgain looks like it's packaged for Debian, so that may be a better option. It sounds like it aims to be somewhat backward-compatible in its flags, so maybe I wouldn't need to do much to switch to it. I'm currently running mp3gain -o -q -s s [file1] [file2] ... and parsing the output.

[derat]

loudgain looks like it's almost a drop-in replacement. The differences I've noticed are:

• loudgain only computes album-wide gain if -a is also passed. Passing -a to mp3gain annoyingly makes it modify the supplied files.
• loudgain prints Album rather than "Album" in the File column.
• loudgain seems to be about twice as slow as mp3gain. On my laptop, two files took 2.987s instead of 1.474s. I guess I should upgrade my computer.

The numbers that are produced are different enough that I'd probably want to reprocess my music collection.

[derat]

If my math is right, it may take about 1.25 days to process my music collection with loudgain, compared to 0.6 days with mp3gain. Probably less in reality, since my code should run multiple processes in parallel.