search

complexlogic / rsgain

A simple, but powerful ReplayGain 2.0 tagging utility
Other
272 stars 20 forks source link

[BUG]: Picard plugin unable to normalize volume #113

Open MRodneyMcKay opened 4 months ago

MRodneyMcKay commented 4 months ago

The Picard plugin was unable to determine the correct replaygain tags for high Res classical music. Foobar worked very nicely though. Foobar downsampled the audio, so only it worked with audible frequencies. Is it possible, that it is missing from this and that is the reason, replaygain failed?

complexlogic commented 4 months ago

Can you explain exactly how it's not working? Are you getting error messages? etc.

Also, if you can upload a sample file, that may help in the diagnosis.

MRodneyMcKay commented 4 months ago

I do not think uploading is such a viable idea, since it would mean uploading 2 copyrighted albums. In picard, i have configured the replaygain 2.0 plugin with the ebu r128 specs, loaded up my whole library and run the plugin. I have set up mpd to use album replaygain tags, and it the collection was normalized at all, the difference in volume was waaay bigger, than it sould have. Here are the tags calculated by rsgain tag name Forthnight Mozart flute replaygain_album_gain -13.39 dB -14.42 dB
replaygain_album_peak 1.413871 1.224665 replaygain_track_gain -12.81 dB -5.00 dB replaygain_track_peak 1.041744 0.910073

Here are the tags calculated by foobar, downsampling, discarding non audible frquencies tag name Fortnight Mozart flute replaygain_album_gain -8,28 dB +0.66 dB
replaygain_album_peak 1.123939 0.994838 replaygain_track_gain -7.79 dB +0.05 dB replaygain_track_peak 1.031543 0.927278

With Foobar's calcuation, the replaygain works as intended, every album is the sameish volume-wise (mainting the volume differences between tracks within the album)

complexlogic commented 4 months ago

I don't understand why the tags have 2 values.

MRodneyMcKay commented 4 months ago

From the whole library, i will show 2 tracks, and the different replaygain values.

This table shows the replygain values generated by rsgain/picard. These values do not work, they do not normalize the value. With this, Taylor swift is considerably louder, than the flute + piano duo.

tag name Taylor Swift Pahud - classical
replaygain_album_gain -13.39 dB -14.42 dB
replaygain_album_peak 1.413871 1.224665
replaygain_track_gain -12.81 dB -5.00 dB
replaygain_track_peak 1.041744 0.910073

This table shows the good values, generated by foobar. These values do actualy normalize the volume

tag name Taylor Swift Pahud - classical
replaygain_album_gain -8,28 dB +0.66 dB
replaygain_album_peak 1.123939 0.994838
replaygain_track_gain -7.79 dB +0.05 dB
replaygain_track_peak 1.031543 0.927278

File size is too big, cannot attach directly, so here is a link: https://drive.google.com/file/d/1P0DTsqhDA347Devo5uFgDd_sXAVRKwxC/view?usp=sharing