Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug (Read 15310 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Hey all,

Just a disclaimer - I'm new to using Mp3Gain and trying to learn more, so please bear with me if I'm misunderstanding the technical aspects of it...

My Mp3 collection is mostly Rock and Metal with some Classical and Electronica mixed in, so there are some major discrepancies in song volume between the genres.  I read some stuff on Mp3Gain and it sounded like the least intrusive and overall best way of normalizing the volume.  I then ran track analysis (and later album analysis) on all of the songs and I figured that 92db would be a pretty good level to compromise.  So when I applied that level of gain to the collection (using Track Gain) and began listening, a lot of songs sound like they're ran through a compressor.  In other words, let's say that the song starts out with an electric guitar riff, no other instruments playing and the volume level sounds like it's at 92db, when the rest of the band kicks in, it sounds as if it "sucks" the volume down to a lower level, similar to how a compressor behaves.  If in the song, it goes to a single instrument again, then it seems to get louder and then sucks it back down when the band kicks in again.

Now I'm sure there's a better technical description of what's going on here, perhaps it's applying gain to the "valleys" in the volume to make it more equal throughout the song (although that's definitely NOT what I want), but I'm kind of in the dark on this and haven't been able to Google anything to help me find an answer.  I just want to know what I'm doing wrong when applying the gain.

Another caveat to this, when I undid all of the gain changes in Mp3Gain and looked at the tag data, I saw that the APE and ID3v2 tags still had the Track and Album gain values in there.  I also noticed that after undoing the gain, they still sound like they're compressed.  Do I need to completely delete all of the ReplayGain and Mp3Gain values out of the tags and then delete the tag itself for it to return to normal?

Thanks in advance for any help that you're willing to throw my way and explain to this newbie what's going on.

praeix

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #1
MP3Gain is just that, a gain. It does not apply any type of sound compression to do so. What you are hearing either is present on the mp3 or the player or hardware that you use to listen to it is doing so.

Said that, since you're playing it at a lower volume, the sensation of the sound being compressed might be stronger.

About mp3gain and replaygain.

Replaygain is an algorithm and a tag format.
As an algorithm, tries to determine the loudness of a song so that a player can adjust the volume to make it similarly loud to others.
As a tag format, it is four tags that determine the song's loudness, the album loudness, and the amplitude peak of the song (used to prevent clipping).

MP3Gain is a program that uses the replaygain algorithm, modifies the MP3 data, and stores information on which gain it applied to allow restoring its value.

Undoing MP3Gain should restore the mp3 data to its original gain. Yet, if you have also replaygain tags, those are probably calculated from the mp3gained data, and so, be incorrect once undoed the Gain.

You could try either remove the repalygain tags, or recalculate them (You can use foobar2000 to do so. I think winamp can also calculate replaygain tags).


Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #2
I generally listen to music in 4 places: my computer at work, my computer at home, my Xbox 360, and my car.  This compressor "suck" issue seems to be present on all of them, but less so on my work computer (I use earbuds to listen to music at work).  On all of my computers I use WinAmp for playback with no audio enhancements like SRS or EQ.  I wouldn't think WinAmp would be responsible for this and over the 360 it does the same thing...

So if it's in the song itself, is it just that the compressor "suck" is imperceptible at it's original volume level without the gain applied?  Most of the volume levels of my songs are in the low 100's/high 90's, so the gain I'm applying is almost always making the songs quieter.  There's only a couple hundred songs in my collection that are below 92db.

It's just such a stark contrast.  It seems like it's taken all of the low volume areas in a given song and brought them up to match the peaks.  Everything I've read says Mp3Gain by nature, doesn't do anything like that, but that's just what it sounds like...

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #3
Can you provide a track snippet original and mp3gain processed so that we have a chance to share your experience?
lame3995o -Q1.7 --lowpass 17

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #4
Undoing MP3Gain should restore the mp3 data to its original gain. Yet, if you have also replaygain tags, those are probably calculated from the mp3gained data, and so, be incorrect once undoed the Gain.

Mp3Gain was written such that undoing the gain also recalculates the tags. There is no need to recalculate them again.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #5
Since MP3Gain applies the same gain-change to the whole file... Most likely, you are just perceiving the volume reduction as compression.  Sometimes when compression is used to make the song "louder", people will describe it as "more dynamic", when of-course the dynamics have actually been reduced and it's just louder.

But actually, if you don't have "Don't clip" selected...  You can get clipping, and clipping is a bad-kind of distorted-compression.   I kind-of doubt that's what's going on.

If you are not allowing clipping, MP3Gain won't "touch" files that would be clipped if boosted (or at least it will limit the amount of boost).

Quote
...I figured that 92db would be a pretty good level to compromise.
The higher your gain setting, the more likely you are to clip (or do nothing).  The default setting of 89dB works pretty well for me.  I mostly play my songs randomly, and just about every time the volume sounds "wrong", I find that ReplayGain was not applied.  (I've been using ReplayGain rather than MP3Gain.)

Quote
...so the gain I'm applying is almost always making the songs quieter.
Right.  That's normal.  Since probably over half of your songs are peak-normalized (volume maximized), how are you going to match the perceived volume?    If you don't want clipping/distortion, you can't boost the peak-normalized songs which happen to sound quiet (just or less loud)...    You have to reduce the loud-sounding songs to bring them down to "average" (and boost a few quiet songs).

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #6
I'll try and get a couple of samples together when I get home and post them so you guys can hear what I'm talking about...or maybe you won't and it only does it on my home computer and 360.  I dunno, but I'll give it a shot.  I'm about ready to give up on normalizing it (which I don't want to do, but I also don't want to have these songs sound the way they do with the "suck" on).

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #7
Quote
Since MP3Gain applies the same gain-change to the whole file... Most likely, you are just perceiving the volume reduction as compression.

Yes this is what I figured.  I just didn't know how else to describe it.  Back in the 90's I remember my parents had a Mark VIII with a sound system in it.  When I drove it, I remember cranking the stereo and it sounded as if the bass was getting sucked in, instead of hitting.  And that was because of the compression on the stereo.  That's what it sounds like and perhaps it's just volume reduction interpreted as compression, but I don't understand how it seems to be applied to parts of the track instead of the entire track, which is what MP3Gain is supposed to do...

Quote
But actually, if you don't have "Don't clip" selected...  You can get clipping, and clipping is a bad-kind of distorted-compression.   I kind-of doubt that's what's going on.

I just checked and I don't have the "Don't clip" option selected.

Quote
Quote
...I figured that 92db would be a pretty good level to compromise.
The higher your gain setting, the more likely you are to clip (or do nothing).  The default setting of 89dB works pretty well for me.  I mostly play my songs randomly, and just about every time the volume sounds "wrong", I find that ReplayGain was not applied.  (I've been using ReplayGain rather than MP3Gain.)

What exactly is the difference between ReplayGain and MP3Gain?  I heard ReplayGain mentioned as an algorithm for calculating the gain, but I also read somewhere that MP3Gain *uses* ReplayGain...

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #8
What exactly is the difference between ReplayGain and MP3Gain?  I heard ReplayGain mentioned as an algorithm for calculating the gain, but I also read somewhere that MP3Gain *uses* ReplayGain...

You are correct that ReplayGain is the algorithm. MP3Gain is the name of a program that implements that algorithm.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #9
OK I was just making 30 second clips of the song (Fireflies by the band Finch) to show the intro of the song how it does the compression "suck".  I first made a clip of the bad file, then I removed the ReplayGain information from the file using foobar2000 (MP3Gain had done nothing to this file except Track Analyze it).  I did this while the file was playing and it instantly bumped the volume up.  I then restarted the song and everything was fine.  The playback was going on in foobar2000, not WinAmp.

So, I think I might have figured it out, but want to make sure.  Does WinAmp use ReplayGain data during playback to alter the volume?  If so, I believe the 360 must also and this could be why I'm hearing this.

I'm still going to make the clips so you can hear the difference, but just wanted to ask this and see where it's going...

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #10
Mp3gain modifies the data (losslessly) to make the sound louder or softer in steps of 1.5 dB. Replaygain does not modify the data but instead adds a tag that specifies how much the player (if it is one that implements replaygain) should adjust the volume. Replaygain does not have the 1.5 dB step sizie limitation.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #11
Mp3gain modifies the data (losslessly) to make the sound louder or softer in steps of 1.5 dB. Replaygain does not modify the data but instead adds a tag that specifies how much the player (if it is one that implements replaygain) should adjust the volume. Replaygain does not have the 1.5 dB step sizie limitation.

But does WinAmp compensate in real-time for ReplayGain?  It certainly seems like it from the last thing I tried.  It would also seem that the 360 does as well.  Also, this brings up the question then, MP3Gain adds those ReplayGain tags, so is there a way to get the MP3Gain changes without it applying the ReplayGain tags?  foobar2000 will remove all of the ReplayGain tags after the fact, but it'd be nice to avoid it altogether...

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #12
No, neither Replaygain or MP3gain alter volume in real time. They alter volume globally, they just turn the volume knob for you before you play the song. Once you play the song, no further volume alteration takes place, it doesn't go up or down at any time.

Edit: please do upload 30 second samples, the same clip, one that didn't go through MP3gain and one that did. The way to do it is to make a 30s sample out of a FLAC file, and encode that as MP3.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #13
Properly coded ReplayGain aware players will apply a constant gain to files which contain RG information. Since it's constant, the dynamics of the song are preserved, unless you run into clipping issues (and even that can be avoided by some players). So any "pumping" you hear is likely due to hardware problems (speakers, amp, etc.), or some normalization DSP/misconfiguration in whatever player you use.

Please provide a step by step breakdown about what is happening, this whole thread is very confusing, since you talk about a lot of different playback softwares, and it's not very clear to me how they are configured and what kind of files you tried to play back and so on.
It's only audiophile if it's inconvenient.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #14
We shouldn't be giving any credence to the idea that something may be broken.

The only true reasons behind the OP's experiences are not comparing before and after at the same volume level (if MP3Gain attenuated a track then the playback volume needs to be reduced by the same amount when the MP3Gain is undone!!!), Fletcher-Munson, specifically the way equal-loudness curves flatten with increasning level, different listening environments and expectation bias.

ReplayGain and MP3Gain do not in any way, shape or form apply compression. While clipping does cause a reduction in dynamic range, this cannot possibly happen if the adjustment being performed is attenuation which is certainly the case with the metal and rock tracks.  I suppose another possibility is that a DRC plugin and palyback gain was configured in foobar2000.  This would have to have been done by the user as it is absolutely not done by default and/or behind the scenes by the player.

Regarding RG tags coexisting with MP3Gain tags, whether the RG tags are right or wrong depends on when they were calculated. If they were calculated and added before MP3Gain made an adjustment then they will be right when the adjustment is undone. If they were calculated and added after MP3Gain made an adjustment then they are only correct for that particular adjustment.  My guess, however, is that separate RG analysis was actually never done and as such RG tags were never written.

The OP should read about RG and MP3Gain from discussions and HA wiki articles before making further bogus claims about DRC or changes to EQ.  Providing samples is not necessary. They are not going to support these claims.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #15
I don't even know what to make of this response.

Quote
The only true reasons behind the OP's experiences are not comparing before and after at the same volume level (if MP3Gain attenuated a track then the playback volume needs to be reduced by the same amount when the MP3Gain is undone!!!), Fletcher-Munson, specifically the way equal-loudness curves flatten with increasning level, different listening environments and expectation bias.

The "not comparing before and after at the same volume level" is confusing.  If I apply the MP3Gain attenuation, then of course it's at a different volume level.  How can you have a "before and after" without any changes in the volume level?

Quote
ReplayGain and MP3Gain do not in any way, shape or form apply compression.

I never said that they do.  I said that I'm still learning about these programs in trying to achieve my goal.  I said that I listen to music generally on 4 devices: my computer at work, my computer at home, my 360, and in my car.  3 of those devices present what appears to me to sound like compression when no compression is present on the playback devices.  I'm trying to find out the reason behind this.  I'm not trying to make any assumptions as to what is causing it, but just describing what I'm observing (e.g. that when I had foobar2000 remove the ReplayGain data, the volume increased as to be expected and no more compression "suck" - this is the part that baffles me).

Quote
I suppose another possibility is that a DRC plugin and palyback gain was configured in foobar2000.  This would have to have been done by the user as it is absolutely not done by default and/or behind the scenes by the player.

Quite literally, nothing has been altered in either WinAmp, foobar2000, or MP3Gain.  WinAmp and foober2000 were downloaded via Ninite by myself a week ago when I reformatted my computer and remain untouched as far as any playback enhancements.  MP3Gain was installed later, but no changes to any of the default setup options were performed by myself.

Quote
Regarding RG tags coexisting with MP3Gain tags, whether the RG tags are right or wrong depends on when they were calculated. If they were calculated and added before MP3Gain made an adjustment then they will be right when the adjustment is undone. If they were calculated and added after MP3Gain made an adjustment then they are only correct for that particular adjustment.  My guess, however, is that separate RG analysis was actually never done and as such RG tags were never written.

I don't know how MP3Gain works in conjunction with ReplayGain or when in the process, the changes occur.  I'm attempting to enjoy learning more about how they work, but am having quite a difficult time with that given your responses.  I ran my mp3's through MP3Gain, doing an analysis on the collection.  When I do the analysis, it adds ReplayGain data to the tags when I view them in WinAmp.  I have NOT ran any gain changes at this point.  When I do apply the gain changes, then listen to the tracks, I can observe, again for lack of a better term, compression "suck" in certain areas of the track.  When I undo these gain changes, the tracks still sound the same as before.  I view the tags in WinAmp and the ReplayGain data and MP3Gain data is still in the APEv2 tag.  In foobar2000, I then selected all of the mp3's, right clicked, went into "Tagging -> MP3 tag types..." and deselected the APEv2 tag thinking it would get rid of the tag and all data.  I then tested playback, same issue.  In WinAmp, the ReplayGain track gain values were still present in the ID3v2 tag.  Then I went back into foobar2000 and right clicked, selected "ReplayGain -> Remove ReplayGain information from files", let it run, tested playback again and everything was back to normal, no compression sounding playback.  That's the entire process, me fumbling through things, testing it out from start to finish, as best as I can describe it.

Quote
The OP should read about RG and MP3Gain from discussions and HA wiki articles before making further bogus claims about DRC or changes to EQ.  Providing samples is not necessary. They are not going to support these claims.

You're right in that I should read more about RG and MP3Gain, but I never once said that there's any changes being made to DRC or EQ!  I said it sounds like it's being run through a compressor.  This could be a matter of interpretation, completely relative to how my own ears are hearing this or maybe it's because there's something in MP3Gain that I'm unaware of or maybe it's the playback equipment or maybe it's WinAmp.  I don't know why this is the case, but I know what I'm hearing.  I'm simply saying this is what it appears to be doing and I'm asking for help if anybody has experienced this before.  I would expect a statement like this from a troll, but not a moderator.

I've described this issue as best I can.  I haven't had time to make the samples yet.  I got sidetracked the other night and now I'm swamped at work.  I'll make the samples, even if greynol says it's pointless.

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #16
The "not comparing before and after at the same volume level" is confusing.  If I apply the MP3Gain attenuation, then of course it's at a different volume level.  How can you have a "before and after" without any changes in the volume level?


To make a valid comparison, you need to first listen to the MP3gain'ed MP3 at a volume you're comfortable with, and then listen to the unprocessed MP3 after lowering the volume by exactly the same amount as the gain reduction applied by MP3gain. Ideally, you would record both outputs, volume match and then use foobar2000's ABX comparator to determine whether you actually hear a difference or not.

I said it sounds like it's being run through a compressor.  This could be a matter of interpretation, completely relative to how my own ears are hearing this


The truth is, we all know here that MP3gain and Replaygain do nothing fancy to processed files. Unless something else is wrong in your playback chain, you're likely simply imagining things because you didn't volume match your comparisons and "louder is better".

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #17
It seems unlikely, but Windows 7 at least can perform Dynamic Range Compression a loudness boost during quiet sections, but it does so globally. It would not apply it only to one file or another, but would do so for all audio.

To check, Right click the speaker icon in System Tray, Select Playback Devices, Select the active device and click Properties, under Enhancements there's a Loudness Equalization, which boosts low levels to louder using DRC. There's also a Disable All Enhancements feature which is by default ticked to send the sound unaltered.

The effects would be rather dramatic, however, so I imagine it's simply a matter of human perception due to the volume change (I've experienced the similar enhancements in perceived quality by turning up the volume linearly by 3 to 6 dB), which would be matched if you applied the same volume change with a volume control.
Dynamic – the artist formerly known as DickD

Attempting To Normalize My Collection Using Mp3Gain, Sounds Ran Throug

Reply #18
Quote
ReplayGain and MP3Gain do not in any way, shape or form apply compression.

I never said that they do.

You certainly put forth the notion that they might.

This was taken from your first post:
So when I applied that level of gain to the collection (using Track Gain) and began listening, a lot of songs sound like they're ran through a compressor.

Whether or not you're claiming MP3Gain or ReplayGain actually apply compression is irrelevant; we are telling you they do not.

I don't know how MP3Gain works in conjunction with ReplayGain or when in the process, the changes occur.

MP3Gain runs ReplayGain analysis but stores the information in APEv2 tags along with additional tags so that the changes it makes can be undone.  This RG information stored and maintained by MP3Gain reflects and gain changes applied by the program.  If separate ReplayGain analysis is performed by some other program and the data is stored in ID3v2 tags then it may or may not conflict with the tags stored in APEv2 tags depending on whether the analysis was performed before MP3Gain was allowed to make changes.  MP3Gain is not aware of what might be stored in ID3v2 tags.  How other programs handle reading and writing RG tags will vary, though they may be able to be configured one way or another.  In my previous post I incorrectly implied that MP3Gain does not write RG tags.  Sorry if that caused any confusion.

I don't know why this is the case, but I know what I'm hearing.

I gave possible reasons but you failed to acknowledge some of them in your unnecessarily defensive response:
Fletcher-Munson, specifically the way equal-loudness curves flatten with increasning level, different listening environments and expectation bias.