1. Back in build 14, there was a serious regression in how album gain was handled. Albums should be leveled using the average gain, rather than using the gain from the loudest track.
This was introduced because, at the time, videos would also use album gain, and were easily driven to clipping.
That's no longer the case for videos, yet we still have this behavior which can cause audibly uneven playback across albums now. One loud track can make everything else on an album too quiet.Addressed in build 120 2. There are a couple of edge-cases which still need to be taken care of:
2a. It doesn't affect me, but a while back, someone requested that the "Radio" Media Sub Type behave the same way that podcasts do, forcing track-based leveling rather than using album-based leveling. This seemed to make a lot of sense.
2b. We also need some way of marking an album as a "Mixtape" to force track-based leveling rather than using album-based leveling, when the files are sourced from multiple albums. One example people used was that they might create a "Top 40" album or a "Greatest Hits" album when they already own all the tracks on that disc.
3. Volume Leveling is performed
before the volume control, which can cause uneven playback if you run out of headroom. The leveling target should be linked to the volume control so that an adjustment of -6 dB reduces the target from -23 LUFS to -29 LUFS for example. This way you gain additional headroom for leveling very dynamic tracks when the volume is reduced. From analyzing my library, I would need at least 7 dB of additional headroom for volume leveling to work well with videos. I don't see any downside to implementing this.
4. Volume leveling is still quite uneven when downmixing multichannel audio to stereo. I only use stereo playback, and I do not foresee that changing, so it would be useful to have some way of performing a downmixed analysis for files.
5. Now that it has been brought up again, perhaps some of the fields should be renamed, and values changed to match industry standard terms? "Dynamic Range (R128)" to "Loudness Range (LRA)" for example.