Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[DoubtingThomas]

I find myself adjusting volume between albums quite often using Volume Leveling, and I don't remember ever feeling that need with Replay Gain...

It would be nice if there was an option to go back to Replay Gain... maybe I'm wrong, but the volume differences between albums with Volume Leveling is annoying in ways I don't remember with Replay Gain.

[6233638]

The issue is not related to the algorithms used for analysis, but how Media Center uses that data.
 
In the earliest builds of MC19, Volume Leveling worked exactly as you would hope - using the average level of an album to adjust the volume, and you rarely ever had to touch the volume control.
 
 
A handful of builds later though, a change was made to use the level of the loudest track in a "group" for leveling, rather than the average.
This was done to work around a problem when volume leveling was used with video playback, as videos were frequently driven to clipping when using the average value in a "group".
 
However, videos are now handled differently (leveled individually, rather than in "groups") so this fix is no longer required.
 
 
But the change from using the loudest track rather than the average volume has not been reversed, and so this problem of uneven volume between different albums persists.
 
I would really like to see this addressed, as volume leveling prior to this change was about as close to perfect as you're going to get, and volume leveling on a track-by-track basis (e.g. using Play Doctor, rather than Album playback) is superb.

[Hendrik]

From what I remember, thats not the reason for the change. The video gain was actually changed one build before this change.

The problem is that if you have a mixture of very quiet and very loud tracks in one album, the average gain could call for a gain which is unachievable by the loud tracks due to their peaks, without running into clipping.
I suppose it could try to be smart, calculate the average gain and then try to check if its a viable value based on the true peak.

However, most audio I have is way over the -23 LUFS target, I wonder what kind of audio is actually below that it would get its volume increased.
Maybe I need more lossless audio, and not bad MP3 versions. :p

Personally, I would implement it to the letter of the spec and normalize everything to -23 LUFS, screw album gain.

[6233638]

I would say that it's primarily Classical and Jazz music which gets positive values, though I do have some Electronic tracks, and Rock/Pop from the '80s where it also applies.
 
These all have Dynamic Range (DR) values in the 11-20 range, rather than today's releases which are lucky to be in the 6-10 range.
 
 
I did analysis on my library a while back and it was something like 0.1% of tracks which would be adjusted loud enough to introduce clipping with a -23 LUFS target if the average level was used.
 
If I recall correctly, there was as much as a 7dB difference between using the average for an album, and the loudest track on an album for leveling.
 
I would much rather that the average value be used, and clipping protection engaged if necessary.
E.g. the average pushes the gain up +10dB, but MC only adjusts it +9dB because one track would be driven to clipping. (keeping in mind that you target -1dB rather than 0dB to prevent inter-sample clipping, as the analysis is not 100% accurate)
 
That 1dB drop from the average will not make much of a difference as far as leveling is concerned, compared to the 7dB difference between using an average and the level for the loudest track.
 
 
As for leveling on a track-by-track basis, this completely breaks gapless albums which crossfade into one another.
 
It also completely eliminates any album dynamics. Some tracks on an album are meant to be softer and quieter than the others, rather than being played at the same volume as everything else.
 
There still needs to be a way to mark an album as being a "mixtape" though for when you have created a compilation from existing tracks in your library. I'm not sure what would be the best way to do that, however. Right now changing the media sub type to "Podcast" achieves that, as podcasts are always leveled on a track-by-track basis.

[dean70]

I have noticed the same thing. It seems the albums that show a higher crest factor fare worse than those that dont  ?

[Hendrik]

I'm testing a change now that will use the average gain again, and then check the peak gain levels against the -1 dBTP target, and reduce if necessary.

Regarding the "mixtapes", how would you suggest to identify them?

[Vocalpoint]

Regarding the "mixtapes", how would you suggest to identify them?

If Album Type was accessible - I would nominate that field for this use. But I suspect it's read-only for a reason.

Cheers!

VP

[mwillems]

I would say that it's primarily Classical and Jazz music which gets positive values, though I do have some Electronic tracks, and Rock/Pop from the '80s where it also applies.
 
These all have Dynamic Range (DR) values in the 11-20 range, rather than today's releases which are lucky to be in the 6-10 range.
 
I did analysis on my library a while back and it was something like 0.1% of tracks which would be adjusted loud enough to introduce clipping with a -23 LUFS target if the average level was used.
 
If I recall correctly, there was as much as a 7dB difference between using the average for an album, and the loudest track on an album for leveling.

Those were similar to my findings as well, primarily classical music needed a positive adjustment, but about 1% of my collection was affected.  The difference between "average" and "loudest-track" based leveling can be exceptionally large on classical albums where the quiet movements might need a positive boost, but the loudest ones need 8dB of attenuation.  Which is all fine, intertrack dynamics are important, but that can create a big difference in volume between albums.  The change in volume between albums based on the "loudest track" method is most significant when going from a very compressed album (which has very uniform leveling needs) to an intermittently quiet classical one (that has widely varying leveling needs).  

I'm glad this is getting some attention, it's one of a very few reasons that I ever reach for the volume anymore.  
 

I'm testing a change now that will use the average gain again, and then check the peak gain levels against the -1 dBTP target, and reduce if necessary.

That's exciting news.  Any implementation based on average gain would be a big improvement for me.  What would make it perfect, from my perspective, is if it took into account additional headroom created by internal volume when determining if the average would "blow over." I.e. if the average recommended a -10dB adjustment, but that would cause one track to go over the target by 2dB, volume leveling would apply -12dB when volume was at max, but if internal volume were at, say, 90%, it would just use the estimated average unaltered (-10dB). But I can imagine that might create some other issues (i.e. what happens when someone increases the volume? Probably clip protection kicks on which creates support concerns, etc.).  

I use internal volume as my primary volume control, so I almost always have at least twenty dB worth of headroom available, and a method that allowed me to always get the exact average (because I have the headroom) would be preferable for me, but I understand completely if that would make things unnecessarily complex.  

Like I said, I'm excited regardless of the specific implementation  

Regarding the "mixtapes", how would you suggest to identify them?

One method that might work is to look for an Album Artist(auto) tag with "multiple artists" or "various artists" type values.  I don't know how other people use album artist, but JRiver's default logic would populate album artist(auto) with "multiple artists" for a mixed "album" like that.   That solution would work for me as I only ever allow album artist to show "multiple artists" when it's a compilation or mixtape, but that might not be an ideal fix for everyone though. Many regular albums have multiple artists on them as well, and if you don't use the album artist tag consistently you could get unexpected behavior; people seem to get hung up on the Album Artist family of tags.

[DoubtingThomas]

I'm testing a change now that will use the average gain again, and then check the peak gain levels against the -1 dBTP target, and reduce if necessary.

Well I'm glad I'm not the only one experiencing volume leveling issues.  Thanks everyone for the input.

Thanks Hendrik for looking into this.

Album Gain is critical.  I always listen to full albums, I never ever play random tracks from random albums and I never ever randomize tracks.

[Quixote]

Regarding the "mixtapes", how would you suggest to identify them?

iTunes has a "Part of a compilation" check mark in the info tab.  I'm not sure you want to be "like" iTunes though. It is how they've implemented compilations though.

The AlbumType field makes a lot of logical sense though.  It's where I would think to look for it.

[6233638]

I'm glad this is getting some attention, it's one of a very few reasons that I ever reach for the volume anymore.  

Definitely.

That's exciting news.  Any implementation based on average gain would be a big improvement for me.  What would make it perfect, from my perspective, is if it took into account additional headroom created by internal volume when determining if the average would "blow over." I.e. if the average recommended a -10dB adjustment, but that would cause one track to go over the target by 2dB, volume leveling would apply -12dB when volume was at max, but if internal volume were at, say, 90%, it would just use the estimated average unaltered (-10dB). But I can imagine that might create some other issues (i.e. what happens when someone increases the volume? Probably clip protection kicks on which creates support concerns, etc.).  

I seem to recall the issue being that Volume Leveling is calculated and applied before anything else, including the internal volume control. So "clipping protection" has already limited the amount of adjustment before the Volume Control is applied.

What I had previously suggested was that volume adjustments change the target level. So at 0dB the target is -23 LUFS, and at -7dB the target would change to -30 LUFS for example.

One method that might work is to look for an Album Artist(auto) tag with "multiple artists" or "various artists" type values.  I don't know how other people use album artist, but JRiver's default logic would populate album artist(auto) with "multiple artists" for a mixed "album" like that.   That solution would work for me as I only ever allow album artist to show "multiple artists" when it's a compilation or mixtape, but that might not be an ideal fix for everyone though. Many regular albums have multiple artists on them as well, and if you don't use the album artist tag consistently you could get unexpected behavior; people seem to get hung up on the Album Artist family of tags.

iTunes has a "Part of a compilation" check mark in the info tab.  I'm not sure you want to be "like" iTunes though. It is how they've implemented compilations though.

Compilation albums are not "mixtapes".
"Mixtapes" would specifically be if I owned all of a band's studio albums, and then pulled the individual tracks from those to recreate a new "Greatest Hits" collection that just came out.
Because the tracks are from separate albums, they are likely mastered to different volume levels, and would need to be adjusted individually.
 
This is not something I personally do, but I've seen people request it quite a few times since Volume Leveling was made to be automatic rather than having to manually switch between album/track-based leveling.
 
Compilation albums would be a "Top 40" album where every track is from a different artist, but the album has already been mastered so that these groups of tracks sound good together.
Many compilation albums crossfade the songs into one another, so you cannot level these individually without completely breaking playback.
 
"Mixtapes" should use track-based leveling, Compilations should continue to use album-based leveling.
 
As for marking an album as being a "mixtape" I'm not sure what the best solution would be.
"Album Type" sounds like the right place, but it looks like a calculated field which should probably not be made editable.
A new boolean field might be best, rather than trying to repurpose something else.

[Fred1]

I'm hearing almost always random tracks from smartlists.
Therefore i very often have to push the volume knob up and down - that's not very convenient.

Besides that, i get very big loudness differences between .mp3 files and .m4a files.
M4a files are always significantly louder than .mp3 files.

[Vocalpoint]

Well I'm glad I'm not the only one experiencing volume leveling issues.  Thanks everyone for the input. Thanks Hendrik for looking into this.

Album Gain is critical.  I always listen to full albums, I never ever play random tracks from random albums and I never ever randomize tracks.

While I am not doubting a possible "issue" here - I have to say since reanalyzing my entire 65000+ FLAC tracks using V19's R128 standards etc - I cannot recall any full albums or any Smartlists (which I use a lot) - needing manual correction by reaching for the volume knob.

Doesn't matter which tracks are in play or what era, mastering or any other possible parameter - all my Smartlists playback at the perfect level - with no odd variance whatsoever.

Is it possible to get a couple of examples of any full albums where you find yourself having to adjust the volume? If an overall change in track levels is by design (As the artist intended) I am not sure I would want software monkeying around with the intended levels.

VP

VP

[mwillems]

Compilation albums are not "mixtapes".
"Mixtapes" would specifically be if I owned all of a band's studio albums, and then pulled the individual tracks from those to recreate a new "Greatest Hits" collection that just came out.
Because the tracks are from separate albums, they are likely mastered to different volume levels, and would need to be adjusted individually.
 
This is not something I personally do, but I've seen people request it quite a few times since Volume Leveling was made to be automatic rather than having to manually switch between album/track-based leveling.
 
Compilation albums would be a "Top 40" album where every track is from a different artist, but the album has already been mastered so that these groups of tracks sound good together.
Many compilation albums crossfade the songs into one another, so you cannot level these individually without completely breaking playback.
 
"Mixtapes" should use track-based leveling, Compilations should continue to use album-based leveling.

I agree that you'd want the album treatment for premixed and faded compilations like you're describing; I meant something entirely different when I mentioned compilations.  I own lots and lots of compilation albums (probably a few hundred), but they don't normally have crossfades.   Mine are mainly either put together by record labels (e.g. best of Sun Records, catalog samplers, etc.), or as retrospectives (e.g. British Punk Music 1978-1981).  For about 99% of those there's no crossfade, and because the tracks are pulled from many different sources, they often have no particular volume relationship with each other (i.e. presented as-is, in chronological order), so I would very much want the volume leveled on a per track basis.  

I do have a very few DJ-set or "mega-mix" type albums, and those do have crossfades, etc., but those are not what I meant when I was talking about compilations, which is my own fault for not being clearer. They're definitely compilations, but in my head mixes like that are a different kettle of fish entirely, mostly because we hated receiving them so much when I worked in radio. They were nearly useless to us on the air because of the baked in cross-fades.  That's probably part of why I don't own very many  

Tl;DR, I was thinking of compilations as basically commercially produced "mix-tapes," because almost all the one's I own are exactly that.  For compilations in the sense you mean that were crossfaded and pre-leveled, you'd obviously want the album treatment for those.  

Still not sure what the best way to iron out the underlying issue is, but a new boolean field is probably the safest thing (it would best satisfy competing needs).

I seem to recall the issue being that Volume Leveling is calculated and applied before anything else, including the internal volume control. So "clipping protection" has already limited the amount of adjustment before the Volume Control is applied.

What I had previously suggested was that volume adjustments change the target level. So at 0dB the target is -23 LUFS, and at -7dB the target would change to -30 LUFS for example.

That would probably work; alternatively, because the volume leveling required is a fixed value in the tags, you could resolve the problem by just processing the two volume adjustments in the opposite order (internal volume first, then volume leveling).  Either way would potentially resolve both the using-the-average-causing-clip-protection problem, and the problem of albums that need positive boost.

[6233638]

I'm hearing almost always random tracks from smartlists.
Therefore i very often have to push the volume knob up and down - that's not very convenient.

Besides that, i get very big loudness differences between .mp3 files and .m4a files.
M4a files are always significantly louder than .mp3 files.

Have you analyzed your library and enabled Volume Leveling? Random tracks should be using track-based leveling already.

I agree that you'd want the album treatment for premixed and faded compilations like you're describing; I meant something entirely different when I mentioned compilations.  I own lots and lots of compilation albums (probably a few hundred), but they don't normally have crossfades.   Mine are mainly either put together by record labels (e.g. best of Sun Records, catalog samplers, etc.), or as retrospectives (e.g. British Punk Music 1978-1981).  For about 99% of those there's no crossfade, and because the tracks are pulled from many different sources, they often have no particular volume relationship with each other (i.e. presented as-is, in chronological order), so I would very much want the volume leveled on a per track basis.

While they may not be crossfading into each other, they are typically at least mastered to be roughly the same volume already.
 
But if you do want track-based leveling, you could still mark those as being a "Mixtape" - which really makes me think the option should not be called "Mixtape" but something like "Always use track-based leveling"

I do have a very few DJ-set or "mega-mix" type albums, and those do have crossfades, etc., but those are not what I meant when I was talking about compilations, which is my own fault for not being clearer. They're definitely compilations, but in my head mixes like that are a different kettle of fish entirely, mostly because we hated receiving them so much when I worked in radio. They were nearly useless to us on the air because of the baked in cross-fades.  That's probably part of why I don't own very many  

Oh I don't own many of them either, but it's still something that needs to be accounted for.

[mwillems]

But if you do want track-based leveling, you could still mark those as being a "Mixtape" - which really makes me think the option should not be called "Mixtape" but something like "Always use track-based leveling"

Yeah basically I'm imagining a "Volume-leveling mode" boolean tag with the true mode set to track, and the false to album.  That way if the tag were unmarked (empty), it would default to album leveling like now (other empty tags tend to show a zero result in expression based searches). 

[Fred1]

Have you analyzed your library and enabled Volume Leveling? Random tracks should be using track-based leveling already.

Yes, of course, i did. But the leveling works very bad for me!

[Vocalpoint]

Yes, of course, i did. But the leveling works very bad for me!

How could it be bad? I guess I must be missing something here

VP

[mwillems]

Fred1, if you look in audio path, do you see volume leveling working there?  Does it show the expected values?  That would be the first step in troubleshooting.

You don't happen to have adaptive volume/peak level normalize turned on do you?  That could defeat volume leveling and create the kind of variations you're experiencing. 

[Vocalpoint]

You don't happen to have adaptive volume/peak level normalize turned on do you?  That could defeat volume leveling and create the kind of variations you're experiencing. 

This is what I was wondering...

VP

[mojave]

You don't happen to have adaptive volume/peak level normalize turned on do you?  That could defeat volume leveling and create the kind of variations you're experiencing.

Just to clarify, that doesn't cause variations in a current Playing Now. For example, if you add three albums to Playing Now and start playback, the volume will be fixed for all albums.

This does cause variations between different Playing Nows and even when adding something to the current Playing Now. If adding the the current Playing Now, it seems like the volume won't change until getting to the album that would be clipping due to peak level normalize.

[Fred1]

You don't happen to have adaptive volume/peak level normalize turned on do you?  That could defeat volume leveling and create the kind of variations you're experiencing.  

Yes, you are right - turned it on now and hope this will resolve my problem.
Thanks!

[mwillems]

Yes, you are right - turned it on now and hope this will resolve my problem.
Thanks!

Just to be clear, I was advising leaving adaptive volume off, not turning it on.  Ideally, if you want the most consistent volume levels, you should have volume leveling enabled in DSP studio, and adaptive volume disabled.

[edited for clarity]

[mojave]

Just to be clear, I was advising leaving adaptive volume off, not turning it on.  Ideally you should have volume leveling enabled in DSP studio, and adaptive volume disabled.

Why do you say "ideally"? Adaptive Volume/Peak Level Normalize doesn't add more volume than is lost with volume leveling. Also, my convolution filters add attenuation that I can't get back. Adaptive Volume/Peak Level Normalize prevents unnecessary attenuation by volume leveling. I can't get enough output sometimes without it.

[mwillems]

Why do you say "ideally"? Adaptive Volume/Peak Level Normalize doesn't add more volume than is lost with volume leveling. Also, my convolution filters add attenuation that I can't get back. Adaptive Volume/Peak Level Normalize prevents unnecessary attenuation by volume leveling. I can't get enough output sometimes without it.

I didn't meant "ideally" in the global sense, use cases vary a lot, and there are obviously cases where peak level normalize is actually ideal (like yours).

Fred1 was having trouble with inconsistent volume levels, and I just meant that turning off peak level normalize would get him the most consistent volume levels without having to touch the dial. I only brought it up, because I've been in two or three threads with folks who thought volume leveling wasn't working correctly because they'd also enabled peak level normalize, and didn't realize that you'd never get even volume levels between albums or playlists that way.

I'll edit my post to make it clearer that those settings are only ideal from a "never touching the volume control" standpoint.

[DoubtingThomas]

 For the record... Adaptive Volume/Peak Level Normalize is all turned off...

[mwillems]

For the record... Adaptive Volume/Peak Level Normalize is all turned off...

Sorry, that "he" in my peak normalization post wasn't you, I was talking about trying to help troubleshoot Fred1's issue, which seemed different from yours [I edited it to make it clearer].  

Based on what you described, I assumed you were experiencing the same issue 6233638 and I have been having, related to changes in the album leveling algorithm when the new standard was introduced (which wouldn't be fixed by turning off adaptive volume).  

[Hendrik]

Coming back to the custom mix albums that want track based leveling..

I don't think Album Type is something good to use, since its automatically handled and that might cause headaches.
On the other hand, I don't really like a on/off boolean option either (we don't even have a boolean type field, so it would end up being an integer, or a list field with 2 possible values)

This turns out to be a rather annoying issue.

[6233638]

I'm not sure what the intended use for Media Sub Type is, and if that might be the correct place for it.
I see that "Remix" and "Single" are some of the options there (great, now I have to consider using the "Single" sub type and changing all my views…) so perhaps "Custom Mix", "Mixtape", or whatever you want to call it belongs in there?
 
We already have special behavior for the Media Sub Type of "Podcast" (and, if I recall correctly, it should also be applied to Radio but is not)

[Hendrik]

Doing it over the Media Sub Type might work, as long as people then don't complain that the commercially produced mixtapes (which may be mastered properly) get the same handling once this sub type is used.
Not that I think there is much inter-track dynamics to be preserved when you have a compilation of random songs from random artists, even if mastered commercially.

[6233638]

Doing it over the Media Sub Type might work, as long as people then don't complain that the commercially produced mixtapes (which may be mastered properly) get the same handling once this sub type is used.
Not that I think there is much inter-track dynamics to be preserved when you have a compilation of random songs from random artists, even if mastered commercially.

No, the main issue is when they crossfade into one another. I doubt most people would even notice if you used track-based leveling on "regular" compilations. (that's not a suggestion to use it by default!)
 
I'm still a bit unsure about using Media Sub Type though. I'd rather that all my music was tagged as such.
But with "Remix" and "Single" being options already, that doesn't seem to be the intention for Media Sub Type.
I'm still a bit unclear on what its intended function is, but "Custom Mix" doesn't seem like it would be out of place among the existing options.
 
Perhaps someone else who has been using Media Center for longer than I have could weigh in on this?