I rip everything to FLAC and I have a mix of very quiet and ear-shattering tracks. Since I play on shuffle, if an ear-shattering track came on at high volume I'd have to give up the music hobby - would replaygain keep the continuous tracks at the same volume (e.g. The Wall)? If it can, and it'll also hopefully make all my music at a comfortable level, that'd be awesome

The answer is yes. And, since the only thing it does to your flac files is adding a few tags, you can try it yourself without getting into trouble

Yay! I'll report back with results. I'd have to use album replaygain, right?

No, I don't think so.

If you search replaygain on google, you'll find the hydrogenaudio wiki page among the first hits. That should explain it.

Depends on what you want.

Album mode adjusts all the tracks by the same amount so that the loudest part of the loudest track has a level of 89dB and the relative levels between all of the tracks stay the same. In track mode, all the tracks are adjusted so that the loudest part is 89dB.

If you're shuffling songs bust still want to keep tracks that are softer than the others in the context of an album soft, I'd go with album mode.

EDIT (after ...Just Elliott's post #6):
If you want those soft album tracks to have the same peak level as all the rest of the tracks, go with track mode.

I did Album gain and it worked PERFECTLY! Thanks everyone for your help!

I might have to adjust it to set the base a bit louder... or I could turn the volume up on my headphones

Edit: Hmm, how DO you change the base level of replaygain? I could have sworn there was a preferences screen...

In windows I use foobar2000 as player, not use preamp for RG'ed files in it's settings but use K(ernel) S(treaming) output - foobar plays directly to soundcard not using windows mixer and setting of PCM volume. And for windows' sounds I set this volume to ~50% (Master volume = 100%). Now actual volume of sound I change only at output amplifier. So I have no data loss from clipping (for lossy codecs pick value often can be > 1, sometimes 1.5 and above). In linux my player is MPD, it also can set preamp value, but now it's set to 0 dB (anyway I turned off system sounds).
My main aim - the best sound w/o clipping, so I have preamp = 0 dB, but you just can set it in your player to +6..+10 dB to have loud sound.

I meant that every track was "normalized" to a bit lower than would be nice. I seem to remember there was an easy way to change the "base".

Just change the value in the Target "Normal" Volume box to the level you prefer. I have changed this setting to 92 dB, which suits me best.

Cheers!

Is this somewhere in Preferences? Because I can't find it.

Please excuse my n00bness

It's in the main window, just below the Add File and Ad Folder buttons.

Umm..?

Sorry ... I was pointing at MP3Gain application, not at Foobar.

In Foobar2000 you have to rightclick the selected tracks, select Replaygain and apply the option you want. First do a replaygain scan and if you want to change the play back level, you can manually edit the track gain afterwards.

edit:
Note that Foobar reports the correction value for track gain that will result in play back level at 89.0 dB. Hence it will show a negative value. If you want to increase the playback level then you have to decrease this value.

Go to the Preferences -> Playback page and setup the Preamp setting.

arpeggio: MP3Gain doesn't do much for flac files

just elliot: You could boost the volume by digitally preamping it (I'm sure foobar has a module for that), but note that this might remove some dynamic range from the recordings, since they'd have to be compressed to avoid digital clipping. The original specification used a 83dB reference level, so the current 89dB is already quite a boost. You may be better off just turning the volume knob up

Just go to Foobar2k, preferences, playback.

On that screen, there's "Replay gain, pre-amp"

Set it as you want. If it's too high, you'll get clipping.

If you tick the "use peak info to scale down tracks that still clip after applying ReplayGain" box you won't get clipping, because the pre-amp setting will be over ridden on a per-song basis to ensure that each song doesn't clip.

If you abuse this (e.g. +12dB pre-amp, no clipping), then almost all songs will end up being peak normalised, and won't sound the same loudness as each other.

Hope this helps.

Cheers,
David.

No, if you apply preamp too aggresively, the files will clip. If you switch clipping prevention on, they will be amplified to the maximum volume possible without clipping.

If you turn the advanced limiter on, then it might compress the tracks, but just the loud bits.

edit: added "too aggresively"

I set it to +3.5dB (92.5dB) but I'm worried about clipping as well as another problem - doesn't this just amplify the RG output? And if so, wouldn't I get better results if I just made the RG values, well, louder? I'm just confused

Manually editing the RG values means you're changing the measured loudness of the tracks. This is not what you want.

Keeping the values unchanged makes sense because the values are measured against a common target volume.

You want the RG values to use a common target volume; if all the files use different targets, it will be impossible to achieve equal loudness with RG.

edit: This doesn't look very logical as a whole, but the individual statements are meant to make sense.

---

If you think all the songs are too loud or too quiet, you can adjust the preamp.

---

If one song is too loud or too quiet, you can edit the RG value of that song to change the measured volume.

Since RG is a standard, I strongly suggest not to change the base value of 89 dB. It is a standard after all.

Better use pre-amp.

Oh, okay then, I'll use preamp - one thing: in foobar, the preamp doesn't seem to affect the spectrum analyzer, so currently the analyzer is very low most of the time. Is there a way around this?

Let me see if I got this straight:

Im only discussing the 'audiophile' replaygain mind you, im not interested in the others. So, from now on please assume replaygaining all files to 0dB (i.e. the loudest peak of the album will be at 0dB, the rest of the peaks will be below that)

- Replaygain will never introduce clipping (and then limit) a file, so no peaks will be hurt in the use of..
- Replaygain will also never result in a LOWER volume then the current file volume, since we're not doing that RMS trick.
- Essentially RG will scan the entire album, find the highest peak, calculate the volume increase to get that peak to 0dB, and then apply that volume increase to all other tracks as well.

My question is this:

Will the resampling that needs to be applied to all tracks not hurt the overall quality of the album more than the often slight improvement of overall volume?

And as a result:

Would it not be very cool to not replaygain albums if they are already within (say) 4dB of maximum anyway? That way SERIOUSLY under-volumed albums will be replaygained, but albums that are already pretty decent will not be hurt by hardly noticable (volume-wise) replaygaining?

Or am I missing some structural point here?

Not trying to be a nitpicker, but is that an accurate statement? I know you weren't trying to mislead the original poster and there are times when I don't always describe things using the right words, but what you said sounds like peak normalization to me, which is the whole reason Replay Gain was invented - to offer a better solution to peak normalizing.

You've just reinvented peak normalization - edit: except it's being done on an album basis. Not very effective.



It doesn't look like peak normalization to me, but I agree it looks weird.

I'm guilty of exactly what I accused him of - not using the right words to describe a situation.

There's no argument on my part that album mode adjusts all the tracks by the same amount or that the relative levels between the tracks stay the same as a result of album mode processing.

What I think was incorrect was saying that "Album mode adjusts all the tracks by the same amount so that the loudest part of the loudest track has a level of 89dB", and "In track mode, all the tracks are adjusted so that the loudest part is 89dB." That's what sounds like peak normalizing to me.

Well no, I recapped (not reinvented) my understanding of replaygain, but my addition is that at a certain point, the loss of actually tinkering with resampling at all will be not worth the 'gain' (i.e. a more even volume across albums).

Hence my suggestion of introducing a certain margin within which replaygain volume compensation will not be applied. More specifically: a LOT of my albums simply don't need replaygain, only a few are too soft overall. I would like replaygain to automatically 'fix' these way-too-soft ones, while not touching most of my albums.

Or is this already a feature of replaygain?

I have trouble understanding you. It looks to me as if you've misunderstood replaygain.

If only a few are too soft, scan just those files and tune the preamp values to your liking. (Assuming you can set preamp separately for scanned and unscanned files.)

I think you've understood me correctly, the only problem is that I do not want to MANUALLY have to go out and find the albums that are too soft. Will take me ages. I would like replaygain to:

- SCAN all my audio files, and write replaygain album data
- ONLY actually do something with the replaygain data if the highest audio peak is below (say) -3dB

So the majority of my albums will NOT be replaygained, only the ones that are WAY too soft.

I still don't understand what tinkering with resampling means. Which leads me to suspect you don't understand RG the same way I do.



You need to stop thinking about highest peaks. RG calculation is a measure of *average* volume, not peak volume. A weighted average.

... Of course you could just replaygain them all and adjust the preamp, i.e. using what it was intended for

Yes, you're missing a structural point. ReplayGain is doing RMS loudness, peaks are only calculated to avoid clipping. It's not 0db (relative volume), but 89db (absolute volume). Read: HA wiki, wikipedia.



OK, you are using foobar, not mp3gain, right? Replaygain is in tags, which can be manipulated like other tags in foobar if you do it right. What you could do is scan & tag your entire collection, then make an autoplaylist using "%__replaygain_album_gain% GREATER -4". Now you can erase the replaygain tags from those tracks.

ReplayGain is there to make all your audio files sound about the same loudness. You can fix the loudness per-track or per-album.

It's not designed to make everything sound as loud as possible.

If you try to make your quieter albums louder, while keeping your loudest albums as loud as they were previously, you'll run into problems, as described here...

http://replaygain.hydrogenaudio.org/faq_quiet.html

Reading all the FAQs might help...

http://replaygain.hydrogenaudio.org/faq.html

Cheers,
David.

Bravo.

Let's end this debate with...

YAY FOR REPLAYGAIN!

Consider my statement as simplistic. I realize that ReplayGain doesn't amplify based on the absolute peak like with normalization but instead determines the loudness by picking the amplitude from the 95th percentile from 50ms sections taking into account that the response of the human ear is not flat.

My point was to illustrate the difference between album mode and track mode and why a person might choose album mode even though the tracks are being shuffled. I think people are too quick to suggest that track mode be used when mixing tracks. Personally, I don't want a track that is quiet on an album be played at an equal volume as a loud track when shuffling.

Must we always have to get bogged down in minutia when trying to explain general concepts? This isn't to say that I couldn't have done a better job however. Perhaps instead of "loudest part" I should have said "loudest perceived part".

YAY FOR REPLAYGAIN!

Ack - that explains it then. I was under the impression that replaygain does not (always) hurt the audio source by limiting and going for 'RMS values' etc.

Guess that makes replaygain useless for me, I really won't allow some program to randomly limit (and therby damage) my sound sources. A pity!

Replaygain is not a limiter.

---

Maybe you should read this website:

http://www.replaygain.org/

It'll take 5 or 10 minutes of your time, but it will save the rest of us hours of trying to explain it piecemeal to you.

HUH?

I hope that isn't in response to what I said.

In case I am in some way still misrepresenting the way RG works, visit the site and read for yourself.

EDIT:
@Mike, you beat me by a minute!

i was reading the link and RG seems very cool(and is of course)
i don't know how to use command lines ,have any GUI where i can adjust the desired level for .wav?

thanks so much.

foobar2000 or one of the frontends for wavgain. A couple of the posts in this very thread explain how.

I think you really need to re-read what ReplayGain is about as I'm not sure you're understanding what it does at all?

It doesn't randomly do anything, nor does it limit anything. Easiest way to see part of what it does...find a quiet album you have and a loud one, now apply replaygain to them using foobar or whatever, then listen to both albums back to back with replaygain turned on and in album mode. They now sound the same volume.

Oh yeah.... YAY FOR REPLAYGAIN!

Not to bog you down in minutia... but the general concept wasn't clear either
The way I understand it, it isn't about the leveling the "loudest part" - which sounds a lot like peak normalization to me, too. I think the algorithm looks at the most prevalent level, not the loudest level or loudest perceived level at all.

And that it doesn't exactly take the most prevalent level (which on many recordings might be silence), but goes down to the 95% line (when all the levels have been sorted by prevalence, if I understand well) - that is, I agree with you, a detail which you may well omit when speaking in general concepts



Hi puntloos, thy seeker for purity
From all your recent topics, I get this feeling that you're afraid of things that "process your sound" in any way... and I know the feeling because that's the same feeling that I used to have. This tool is not doing any evil stuff; it just applies some gain, and may protect you from occasional digital clipping by smoothing out the peaks that go over full-scale after applying a positive gain.

off-topic: in fact, with your high-spec audio hardware, I would suggest you look into "processing" your sound a bit more: www.duffroomcorrection.com

Stop it!

Cheers,
David.

No, the player using the replaygain data will need to use a limiter. That, or do it like i suggested, which is amplify the album to make the highest album peak exactly hit 0dB, no more.

Ah, good advice, after you read it yourself you'll perhaps realise your derogatory remarks only made you look silly.



Hey patsoe, you're right .. my philosophy basically is: try to create an audio chain that alters the original 'waveform' (if you will) as little as possible. Adding noise/coloring or even this DRC stuff is perfectly acceptable, but that should be added by choice. Thats why I don't like tube amplifiers either, they add coloration that sometimes will make stuff sound 'sweeter' but just as often over-sweeten stuff when a recording enigineer decided to already add a bit of sound-fluff.

Thanks for the link btw, I was aware of the technology but maybe doing it in my setup might actually be doable and a worthwhile experiment.

Now back to the original discussion. Let me recap my point of view REALLY quickly:

- Replaygain determines the RMS value of tracks, or albums. This in itself is OK. Do what you will.

The thing I have a problem with is this:

IF an album comes out below the goal RMS (say the album RMS is -20dB and the goal RMS is -10dB), then the replaygain player will 'turn up the volume'. Great, until you realise the sporadic peaks would be amplified over 0dB, i.e. clipping. Oops, well then hard limiting is engaged, compressing that specific peak. To me, this compression is totally unacceptable. One can debate if, and how well you can hear this effect, but call me an audiophile, Id prefer it not to happen if I can help it.

Or, as the replaygain site says, turn down the playback level ('goal RMS setting' if you will). And this is specifically why I would want, as Ive indicated from the start, replaygain to ONLY do peak normalisation upwards, and ONLY if a significant gain can be made. If an album's PEAK level doesn't come above -20, then by all means, make that peak 0dB. But if the album's peak is at -0.0000001dB, then less math is better. Hence my suggestion to make a 'futility limit'.

As for albums that are (say) at -5dB RMS, i.e. way too loud, then sure, turn them down a bit, the damage is probably done already, no need to unnecessarily scare me after playing a track off a softer album

All in all, I think replaygain could be a useful tool to improve the volume levels of albums by turning down the too-loud ones, and turning up the too-soft ones if possible without ANY clipping/limiting. The only way to do this is as I described, I think.

Well, I suppose you might notice the loss of dynamic range in extreme cases.
You can get the behaviour you want by setting the software's preamp such that there's no limiting needed in any case. But then you may be losing resolution because most albums will be played so much quieter...

Indeed. I am very very keen on eliminating any distortion unless absolutely necessary, or strictly justifiable. I would like all my CDs to 'perceptually' sound equally loud as much as the next guy but Im not willing to sacrifice any quality for it. I would therefore opt for the best approximation/improvement path. Hence my other topic: Perfect Volume Control?.. from this topic Ive 'decided' that I wouldnt mind some volume tweaking, especially when it's done in the 24bit domain.

Incidentally it turns out Windows Vista has the DRC stuff built in.. interesting..

puntloos - I think you're being needlessly exact about this. If you want exactly the original sound of the band, why not go to an unplugged live gig with just you in the audiance? Seriously, there's nothing wrong with some modification.

Heh, well, that might be a different discussion. Why do some people buy $500/bottle wine while others feel that $5 wine is just fine.... to each their own. In my case I am simply making an effort keep the audio as untouched as I possibly can. (short of actually going to concerts, obviously, I do that too but I dont think that's a relevant here)

Not only that but I think puntloos is assuming that replay gain results in tracks getting amplified. Most of the time they get attenuated (even with the reference increased from 83dB to 89dB).

@Patsoe, my nitpicking friend :
The 95th percentile may very well be in the loudest perceived part. It depends on how dymanic the music is. For compressed music, you are right, the 95th percentile may exist throughout the track. HOWEVER, the perceived loudest part does not mean the same thing as peak level, no matter how you want to spin it. AGAIN, my point was the difference between album mode and track mode and that track mode isn't necessarily the best solution for shuffling or a compilation. Perhaps the term "perceived volume" or "perceived loudness" instead of "loudest perceived part" will finally put eveyone's mind at ease? I'm still guessing not.

Hehe, actually I thought for once I wasn't nitpicking

Please review my previous post, the key difference compared with your statement is in the word "prevalence".

Anyway, I see the point of using audiophile/album mode

edit - just to clarify, we're still bickering (lol) about this line:

and the "proposed amendment" would get something like this instead:

...where "perceived" says that on tracks with a lot of silent bits in it, those are disregarded...