Print

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

[Matt]

Replay Gain breaks a song into chunks, analyzes their volume, sorts those chunks by volume, then says "hey, the 95% loudest chunk is the reference for that file."  Then, during playback, you compare the reference marks of the different files and adjust the volumes accordingly.

Album Gain does the same thing, except it thinks of albums as one giant song.  That means it takes the 95% loudest chunk of the whole album for the reference mark.

Using Album Gain makes it so albums that have tracks that are supposed to be quiet actually are quiet.

Both types of gains have their place.  Currently MC doesn't support Album Gain.

The solution I'm proposing is for MC to calculate the Album Gain from the Replay Gain of the tracks on the albums.  This requires no extra analysis and avoids the GUI pitfalls of forcing full albums to analyze together.

The Album Gain could be calculated as the loudest track Replay Gain from all the tracks on the album.  A more elaborate calculation scheme could also be employed. (weighted averaging based on volume and duration, etc.)

It seems to me like this would work equally well as calculating the files again.  Replay Gain isn't an exact science, and the difference between the two methods would only be a few tenths of a decimal.  Neither method would really be more correct -- just different.

So, would this make those pining for Album Gain completely happy?

Thanks for any feedback.

[Madcow]

Matt, that would make me very happy and seems like the most sensible and logical solution.  It would mean that each track on an album played at the same relative volume to each other while also ensuring that, overall, all songs played at a roughly equivalent level.

Of course, you'd have to fix the iPod support before I got any real benefit out of it.  

[Nolonemo]

That would make me happy.  I suppose you would select to use either "Relay Gain" or "Album Gain" (or nothing?) for a playback session or playlist.

[Madcow]

Oh yes - that's a good point - for this to be truly of benefit, you'd need to be able to select either "normal replay gain" or "album replay gain" (or "none") in the "Upload Volume Adjustment" box for the portables.  I don't tend to use Replay Gain when listening on the PC, but it's absolutely essential to my listening pleasure on the move.

[lee269]

Matt, that seems sensible to me. Some sort of average would I think definitely be preferable, and mitigate the effects of situations where one track in an album is particularly loud or 'incorrectly' assigned, as described in this thread. Not sure how common this problem is though.

Getting OT for this thread, the ability to adjust RG values manually would also solve this particular problem.

[Matt]

Matt, that seems sensible to me. Some sort of average would I think definitely be preferable, and mitigate the effects of situations where one track in an album is particularly loud or 'incorrectly' assigned

If you based it off the loudest track (the simple way suggested above), extra quiet songs wouldn't matter.

The problem would be if there were a CD with lots of quiet tracks and one loud track.  It'd make the CD play quieter by comparison to other CDs most of the time.  I've never seen a CD like this, and even if there were some oddballs, it'd throw off both album gain methods.

[bspachman]

The solution I'm proposing is for MC to calculate the Album Gain from the Replay Gain of the tracks on the albums.  This requires no extra analysis and avoids the GUI pitfalls of forcing full albums to analyze together.

The Album Gain could be calculated as the loudest track Replay Gain from all the tracks on the album.  A more elaborate calculation scheme could also be employed. (weighted averaging based on volume and duration, etc.)

It seems to me like this would work equally well as calculating the files again.  Replay Gain isn't an exact science, and the difference between the two methods would only be a few tenths of a decimal.  Neither method would really be more correct -- just different.

Sounds plausible enough in theory. In practice...what happens when you aren't listening to albums in order?

For example, if you have a Playing Now that consists only of your 4-5 rated tracks, ReplayGain being enabled will generally level out the volume differences between those songs.

What if you enable AlbumGain instead? Will the algorithm only take into account the tracks of the various albums already in PN? Will it take into account the other tracks from the various albums (the tracks not in PN?)

How about redoing this feature as "Gain Leveler" (or something). If you enable the GainLeveler, it first takes into account all complete albums in PN, then for non-complete albums, it uses individual track values to come up to the album level. That should make the orphan tracks as loud as the loudest album track, while preserving internal album differences.

Of course, there will be folks who will want to override the complete album behavior, so there would need to be some way to force the current ReplayGain behavior even if there are complete albums in PN.

Shuffle mode is another thing to consider...would putting MC into shuffle mode automatically force the current RG behavior? Should AlbumGain have anything to do with shuffle mode at all?

Thoughts?
Brad

[Matt]

Hey Brad,

MC has an album analyzer that knows what album a file is in.  It'll use that collection of tracks when calculating an Album Gain.

What files are or aren't in Playing Now won't matter in that stage.  Does that simplify anything?

Thanks.

(p.s. If you pick "Automatic" in the Replay Gain DSP Studio dialog, that would still look at Playing Now.  This is so if you had only two tracks both at -10dB, it'd just say "what's the point -- use -0dB instead."  This stage works the same whether you're using Replay Gain or Album Gain)

[lee269]

Matt, I think I may have misunderstood.

If you based it off the loudest track (the simple way suggested above), extra quiet songs wouldn't matter

The problem would be if there were a CD with lots of quiet tracks and one loud track.

The particular problem I described (and granted its obviously an unusual/erroneous RG) is one track at +6db abnd the rest at around -6db.

The Album Gain could be calculated as the loudest track Replay Gain from all the tracks on the album.

If you just took the loudest RG in this case the whole album would be approx +6db, right? Out of 4047 music files I have only 17 with positive RG values. Id have thought under the 'max' MC album gain situation this album would therefore be much louder than the rest. Of course my situation with this album (1 out of 100s RG'ed fine) is probably an RG glitch which I currently cannot fix.

[hit_ny]

Of course, there will be folks who will want to override the complete album behavior, so there would need to be some way to force the current ReplayGain behavior even if there are complete albums in PN.

Shuffle mode is another thing to consider...would putting MC into shuffle mode automatically force the current RG behavior? Should AlbumGain have anything to do with shuffle mode at all?

This is the point i wanted to get into. I am hoping there will be more openess in the SDK so Nila could take advantage of this more with his equaliser plugin.  

Specfically  if i were listening to tracks

- in shuffle mode as is very often, i usually choose replay-gain and cross-fade(smooth).
- in album mode, replay-gain is off, (album gain in effect) and gapless is selected.

 

[pbreet]

Matt,

Sounds good.  What do you see as to how we'd choose between album and replay gain?

[Matt]

The Album Gain could be calculated as the loudest track Replay Gain from all the tracks on the album.

If you just took the loudest RG in this case the whole album would be approx +6db, right? Out of 4047 music files I have only 17 with positive RG values. Id have thought under the 'max' MC album gain situation this album would therefore be much louder than the rest. Of course my situation with this album (1 out of 100s RG'ed fine) is probably an RG glitch which I currently cannot fix.

Loud files have lower (negative) replay gains.  Positive replay gain means a loud track.  It's confusing.

[bspachman]

MC has an album analyzer that knows what album a file is in.  It'll use that collection of tracks when calculating an Album Gain.

What files are or aren't in Playing Now won't matter in that stage.  Does that simplify anything?

I get what you're saying from an initial calculation standpoint (the setup, as it were).

My thoughts were based on potential playback situations. Have a mix of full albums, partial albums, and shuffle mode could make it confusing as to how to adjust the gain on a given track.

I love the idea, I hope you can make it as intuitive as possible.

Best,
Brad

[rocketsauce]

My question also would be about how we would switch between "Track Gain" and "Album Gain". Would it somehow be able to be automatic or would we have to change modes manually? If we would have to manually switch from one to the other, then I don't really see how that would be much different than the way MC is now. If I'm playing in shuffle mode, I enable RG. If I'm playing an album or albums (in their proper order), then I disable RG.

Rob

[lee269]

Loud files have lower (negative) replay gains.  Positive replay gain means a loud track.  It's confusing.

Of course youre right Matt - I was referring to the playback effect with RG enabled. In practice in my specific case though the transition from RG+6db to RG-6db  means one track is room-shakingly loud whilst the next is at 'normal' volume. And to make that album +6db (as max RG value) compared to every other album being -ve would induce the same effect across albums, I guess.

Anyway, I think Im beginning to pollute this thread with a one off situation. Any chance of allowing manual RG adjustments ?

[kiwi]

The solution I'm proposing is for MC to calculate the Album Gain from the Replay Gain of the tracks on the albums.  This requires no extra analysis and avoids the GUI pitfalls of forcing full albums to analyze together.

The Album Gain could be calculated as the loudest track Replay Gain from all the tracks on the album.  A more elaborate calculation scheme could also be employed. (weighted averaging based on volume and duration, etc.)

This would definitely work for me (taking the loudest track, as calculated by album analysis and apply that to all tracks in the album).  I say keep it as simple  as possible.  

So, would this make those pining for Album Gain completely happy?

YES!

[xen-uno]

Not to sound like a broken record, but I would like to see MC phase out it's proprietary method of tagging (I am assuming that it still is) in favor of the compatible tagnames as provided by MP3Gain/VorbisGain/etc.

As far as the nuts & bolts of calculating Album (aka Audiophile) gain, Matt's alternative method (based off a collective Radio gain average) sounds great for pre-analyzed files...but what if they haven't been yet? This is where an option (yes I know...another option) should be present to pick either Album or Radio style RG...as done traditionally by the xxxGain progs mentioned above and specified by David Robinson.

V10 is the perfect starting point for complying with some standard...and what Robinson proposes is the closest there is to one.

10^-27

[hit_ny]

My question also would be about how we would switch between "Track Gain" and "Album Gain". Would it somehow be able to be automatic or would we have to change modes manually?

One solution to this would depend on the order of tracks in playing now.

If there were a bunch of tracks in playing now.

- by looking at the album, its possible to see whether the tracks belong to one album or not. If so then apply album gain for those tracks ONLY ( if all album tracks are present in sequence) otherwise replay gain. This was the idea to be used with Nila's equaliser plugin.

As far as i understand album gain, it sets the gain to a certain level (-83dB or -92dB). All other tracks are gained to that level keeping the differences between the album tracks.

[salsbst1]

I like the weighted average idea.  BTW, this solution looks familiar: http://yabb.jriver.com/interact/index.php?board=6;action=display;threadid=15596;start=msg106191#msg106191

[Matt]

That's funny Stuart.  Reminds me of the Seinfeld where Elaine draws a comic for the New Yorker only to find out later it was a complete rip-off of a Ziggy cartoon she had once seen.  Guess that's me.

[salsbst1]

No worries!  I've not been following the development of v10....  Have you chosen a strategy?  has one been implemented?

[paulr]

How's the progress on this going?  I thought I'd ask and give some suggestions...

1.  Use the Replay Gain that MC already calculates and apply a time weighted average to the entire album, or analyze the entire album.
2.  If either Replay Gain or Album Gain is selected and none are available, use a user defined value for attenuation.  (A typical value might be -10dB)
3.  When in "Album Gain" or "Audiophile Gain" (I think that's what the original proposal called it) mode, use Album Gain if available, if not use Replay Gain (radio).
4.  When in Replay Gain (radio) mode, use Replay Gain (radio) if available, if it's not available, use Album Gain.

I am currently using Foobar to calculate Album Gain, and Foobar's masstagger to change MC's "Replay Gain (radio)" tag in order to achieve this result.  I'd much rather have this built into MC.

Another suggestion is to add a control to the menu bar that allows easy switching between these modes.

[Monkeyboy]

Another very happy person
Rich

[GHammer]

I like the idea.
I also agree with those who say they'd like to have MC tag replaygain info as MP3Gain, foobar, and others do.

[Richard Berg]

Sounds good to me; as you say, RPG is not an exact science.  I would like the data stored in some sort of standardized format, though, in case my files are ever used on another player (e.g. a friend who hasn't bought MC).

[UdoS]

Hi,

Album gain only makes sense on an album, where all thrack belong together (e.g. some classical or The Wall from PF). There is no way to fullfill all needs with one setting. Listening to an album (eg the wall) I adjust the volume as I need it without replay gain.
If you listen to ablums containing indiviual songs, there is no need for album gain, rp-gain is fine.
The last question now is: is there any need for ablum gain at all. I f yes, there is only one logical solution: using the loudest track info and apply it to the entire album.

Udo

[paulr]

Album gain only makes sense on an album, where all thrack belong together

Not necessarily.  Some albums do a good job of presenting a 'mood' by varying the loudness and tempo of individual songs.

There is no way to fullfill all needs with one setting.

Yes, you make it optional...  However, this is irrelevent because we are talking about two settings anyway.

If you listen to ablums containing indiviual songs, there is no need for album gain, rp-gain is fine.

In your opinion, maybe.  But there are reasons that make it worthwhile for many of us.

there is only one logical solution: using the loudest track info and apply it to the entire album.

I don't know how this is the *only logical solution*.  It is *a* solution: the easiest and least accurate solution.  It's certainly not the best solution (which would be to analyze the entire album).

Anyway, I was just hoping for some insight on where we are with this feature.  Are we looking at it for version 10.1?  11?

[Fixer]

Matt, that seems sensible to me. Some sort of average would I think definitely be preferable, and mitigate the effects of situations where one track in an album is particularly loud or 'incorrectly' assigned, as described in this thread. Not sure how common this problem is though.

It's very common on live albums or theme albums, like Yes or classical music.

Getting OT for this thread, the ability to adjust RG values manually would also solve this particular problem.

If Album Gain works, user editing would be less useful, but I still agree, it would allow user tweaks.

[Fixer]

Matt, that seems sensible to me. Some sort of average would I think definitely be preferable, and mitigate the effects of situations where one track in an album is particularly loud or 'incorrectly' assigned

If you based it off the loudest track (the simple way suggested above), extra quiet songs wouldn't matter.

The problem would be if there were a CD with lots of quiet tracks and one loud track.  It'd make the CD play quieter by comparison to other CDs most of the time.  I've never seen a CD like this, and even if there were some oddballs, it'd throw off both album gain methods.

I've never seen one either, but making the replay gain field user editable would allow us to fix those one in a million odd ball albums.

[Fixer]

Matt, yes I think your calculation idea may be quite useful.

It brings up questions about how to choose between album or radio gain.  On the other hand, maybe there is nothing to choose.

Suppose I select a set of tracks to do analysis on.  There might be a set of radio buttons to tell MC how to do the calculations.  If Radio Gain is chosen, it works like it always has.  If Album Gain is chosen, it analyses ALL the selected tracks, and then sets the Replay Gain setting for all the selected tracks.  The choice for Album or Radio gain is set for that set of tracks at ANALYSIS time.

What I've described above would work perfectly for me.  I have read other wondering how Album Gain would affect shuffle and incomplete album playback.  Somehow I think it's a moot point.  A few db here and a few db there is not all that important, unless it radically changes between tracks from the same album that should sound contigious.  

I think WE should choose what we consider the tracks of an album to be.    For example, what if there is a 2 disk set that is a double live album and the Replay Gain should apply across the disks.  If our tagging has them as Disk1 or Disk2, it might not come out right, unless I can select all the tracks from both disks and instruct MC to do the album gain for ALL the tracks as an Album.

Anyway, that's my opinion, thanks for keeping the issue alive and thinking of solutions.

Good luck on the user interface.

[UdoS]

to Paulr (or anybody else)

---> But there are reasons that make it worthwhile for many of us.

Can you give some example for use of album gain? (Just for my understanding)

Udo

[Fixer]

Are you familiar with Pink Floyd / Dark Side of the Moon ?

The tracks on that album all run into each other, there is no audible gap between any track.  When replay gain is done in MC it only looks at each track to calculate the value for THAT track.  One track may be -6db, another may be +4db, an infinate set of possible combinations based on the content of the album.

Now imaging listening Dark Side of the Moon, the first track is set for +3db and the next is -8db (I'm making these values up), the volume would drop considerably when switching from track 1 to track 2.  Not only is this audibly disturbing, it is not how the performers intended it to be heard.

Album Gain retains the relative volume BETWEEN tracks of an album.

Think about live or classical music, the volume jumps between tracks, it doesn't sound good.  So without Album Gain, Replay gains only solves some of the volume differences between different tracks SOME of the time and makes it worse SOME of the time.  I believe Album Gain will solve it 99% of the time.

[UdoS]

I just check this album (Dark side of the moon). All values are about -6db. If the ablum is recorded properly, thats the way it should be. Specialy in classical there is very often a high (and wanted) differnce between tracks.

Let me write my understanding about replay gain:

Since there is no standart output volume (as it is in cinema systems) replay gain uses a volume refernce (rms-power) of -20db, in humen beeing is 1% output power) as reference. On a calibrated system, this gives you 20db overhead between the average volume and peak output. These values refer to power, so talking in drive voltages, it would be -10db. This is the value replay gain uses as refernce rms-volume.

Listen to an album (eg. Dark side of the moon) this means: using replay gain, the entire album is played back at a volume equal (playback volume - 6db), if rpg is enabled. Looking at the calculated values, they all were in the range of -5.4 to -6.3 db. So, usíng a calibrated system and reducing the output volume by -6db would have the same effect.

Listen to recordings ripped from differned albums behave totally different. Some companies like to make a powerfull sond by compressing the sound and normalizing it to 0db. This gives you a very high rms-power in comparison to other recordings. Replay gain will now compensate for this almost perfect.

As a conclusion to this, album again should only set the entire album, not an single track, since what you hear then, is not what the artist wanted. Specially Pink Floyd were using loud an silent parts as speical effects in their albums. If you like all tracks play on same volume, the use of a dynamic compressor would be better.

But, since the entire album is adjusted by the same value, listing to an album, turning rpg off and readjust the amplifier should give an equal result. That's why I stated earlier, I see no real need for album gain.

Udo
 

[kiwi]

But, since the entire album is adjusted by the same value, listing to an album, turning rpg off and readjust the amplifier should give an equal result. That's why I stated earlier, I see no real need for album gain.

Yeah, that will work, however, I'd rather not have to get up and change the volume while I'm listening to music.  I have a random albums playlist that's great.  W/ album gain, I can just set it and forget it.

kiwi

[paulr]

Replay gain calculations for Dark Side of the Moon:
1.  Speak to Me/Breathe: -6.51 dB
2.  On the Run:  -4.72 dB
3.  Time:  -9.71 dB
4.  Great Gig in the Sky:  -8.86 dB
5.  Money:  -11.57 dB
6.  Us and Them:  -8.15 dB
7.  Any Colour You Like:  -8.47 dB
8.  Brain Damage:  -8.14 dB
9:  Eclipse:  -11.17 dB

As you can see, the replay gain for each track varies wildly.  There are only two ways to account for this - turn OFF replay gain entirely, or use Album Gain.

Those of us who tend to listen to random albums more often than random tracks need Album Gain in order to prevent the volume adjustments between each album (we want to preserve the relative loudness of each song *within* each album, but 'level' the overall loudness of the albums to each other.  Album Gain does this for us.

I just check this album (Dark side of the moon). All values are about -6db. If the ablum is recorded properly, thats the way it should be.

I'm not sure how you calculated replay gains for Dark Side, but the only way I can see this happening is if you ripped it as one file instead of individual tracks.  MY calculated Album Gain for Dark Side is about -9.69 dB.

[UdoS]

Hi,

I see your point, but I'm not sure, there will be a algorithm to take care of this. Since you're interested in album gain (means you like to keep the relation between loud and silent tracks), I see no problem to set the gain for one single album. But playing more albums (probably different kind of music), bringing those albums together looks to me like comparing apples and bananas (e.g: listening to Pink Floyd The Wall, The  Rolling Stones and some blues artists. You're looking for an algorithm setting up all albums to keep 1. the special touch indicated by loudness and silence and 2. the rms-volume should stay constant.
I think, you have to look for some compromise between different albums. One way I see is to use the already calculated values for rpg and average those for ablum gain. In this case, you don't need to reanalyze all tracks, just calculate the gain value on the fly (by using all albums from the current play list) and use this instead rpg. This could easy be handled by software.

Udo

[Roar]

I just checked the replay gain values on my Ultradisc II Remastering (Mobile Fidelity) version of Pink Floyd - The Dark Side of the Moon.

There is a 19.5 dB difference between the highest and lowest replay gain value!!!

1.	Speak to Me	(+9.23 dB)
2.	Breathe	(-4.31 dB)
3.	On the Run	(-2.27 dB)
4.	Time	(-8.12 dB)
5.	The Great Gig in the Sky	(-6.42 dB)
6.	Money	(-9.91 dB)
7.	Us and Them	(-6.31 dB)
8.	Any Colour You Like	(-7.22 dB)
9.	Brain Damage	(-6.56 dB)
10.	Eclipse	(-10.30 dB)

There's a 14.5 dB drop in volume when going from track #1 to track #2 with normal replay gain. This is the reason that I'm currently running with replay gain switched off in MC!

I don't think that replay gain is an exact science, partly because I don't think that the calculated statistical volume for a track is always what the listener perceives as the subjective volume.

I think that calculating an album gain as a time-weighted average of the already calculated replay gains is "close enough" to an optimal way of doing album gain, and it would be a huge improvement over normal replay gain for those of us, who listen to albums of this kind.

I think that Album Gain could be an alternative to Replay Gain in the DSP options, and that Album Gain can be calculated during Audio Analysis from the individual Replay Gains; for tracks that have already been analyzed, a new Audio Analysis would then not have to re-analyze the audio file, but could just calculate the missing Album Gain.

Other than my comments above, I think that the approach described by "paulr" would be fine.

- Roar

[lee269]

Roar - interesting. I posted about DSOTM a while ago because my version recorded from my vinyl original goes from +6.12 to -7.53 between Speak to Me and Breathe (ripped as individual files). What is the Ultradisc etc version? Id assumed it was a quirk of my recording alone that exhibited this odd RG. The responses I got at the time from people (who had ripped from the standard CD I think) didnt show the same unlistenable RG step between tracks 1 and 2.

Im amazed that someone else has had the same problem with the same album.

And, yes, Id agree that an album average RG would be useful, or the ability to manually adjust RG values in MC to cover odd situations like this. To be fair to MC though, RG works fine most of the time as it is for me.

[EDIT - I just realised I had posted about this in this thread back in January. Anyway, its still interesting to me. And I still havent re-ripped Speak To Me like I promised...]

[GHammer]

Here's what foobar says about the replaygain for DSOTM. Why foobar? It is free and does both track and album replaygain.

T1-
bitrate = 752
samplerate = 44100
channels = 2
bitspersample = 16
flags = 0
codec = Monkey's Audio
compression = Normal
version = 3.99
md5 hash = a559e521cb2c9c7220caed14d55d3dd3
referenced_file = Dark Side Of The Moon.ape
index = 18816
replaygain_track_gain = -0.47 dB
replaygain_track_peak = 0.679779
replaygain_album_gain = -3.68 dB
replaygain_album_peak = 0.968200
 ----------
10606344 samples @ 44100Hz

T2-
bitrate = 752
samplerate = 44100
channels = 2
bitspersample = 16
flags = 0
codec = Monkey's Audio
compression = Normal
version = 3.99
md5 hash = a559e521cb2c9c7220caed14d55d3dd3
referenced_file = Dark Side Of The Moon.ape
replaygain_track_gain = +1.22 dB
replaygain_track_peak = 0.691680
replaygain_album_gain = -3.68 dB
replaygain_album_peak = 0.968200
 ----------
9393300 samples @ 44100Hz

[Fixer]

This is a good discussion.  I'll bet Album Replay gain could be implemented and nobody would notice the difference, except the benefit on Dark Side of the Moon and similar albums.

But hopefully MC will come up with a user interface that allows choosing.

For me, I select a few tracks for this album, maybe all that album, another few full albums, then a few tracks, and I have my playlist for the day.  I want to use Replay Gain.  I do not want to ever have to readjust the volume between albums.  For me I'd set Album Replay Gain on ALL the time.

[Roar]

Slightly off-topic, in response to lee269's question about the Ultradisc II Mobile Fidelity version of Dark Side of the Moon:

Mobile Fidelity Sound Lab (www.mofi.com) has released remastered versions of many well-known albums with excellent sound quality. When I bought the Ultradisc II CD version of DSotM (years ago) I wondered how much better than the normal version of the CD it would be - after exactly ONE SECOND (I'm not kidding) of listening to one of the tracks, I already knew how much better it was; it cost twice the price of the normal version, but it was well worth it!

If you have a decent Hi-Fi system and cherish a good listening experience, and you come across an Ultradisc version of an album you like, I suggest you try it out - I don't think you'll be disappointed.

As to the huge step in replay gain from track #1 to track #2 on DSotM that lee269 and myself experience unlike some others posting here, the reason is that on some CD releases, "Speak to Me" and "Breathe" are combined as one track, whereas on the MFSL Ultradisc 2 (and perhaps other releases?) they are separate tracks.

- Roar