What is best recommended for normalizing Wav files? I am not to big on EAC for normalizing and every time I use Mp3gain I get errors even when I have no check for Layer I and II.

-Darin

:alien:

WaveGain.

Note: It's not lossless. If you want lossless ReplayGain, compress it to FLAC first, then use FLAC's implementation.

I just started reading about Replaygain and I'm really impressed! If I compress my Wav files to FLAC, doesn't that(or any compression for that matter) cause some loss on the original Wav file?

None at all. Try it yourself: Take a WAV, compress it with FLAC, decompress it, do a CRC check against the original. Isn't it beautiful?

It's an advanced ZIP for audio files, so to speak. Don't confuse it with lossy codecs like MP3; that's something completely different.

i never understood this but i knew that normalization is not lossless.
whats the difference between normalization and replaygain. is replaygain good for preventing clipping ? but when a file (cd) is already clipped you can't declip it, can you ?

I would answer your question, but I've been hypnotized by your avatar....




replay gain is a method for analysing the loudness of an audio file, and storing it in the file as metadata
mp3gain takes the replay gain value, and applies a gain change to the mp3 to make the audio in the file match a particular loudness

These two are lossless because
a.) replay gain doesn't change anything about the audio - it's just a value in the header/footer of the file
b.) mp3gain changes something called the "global gain" in each frame of the mp3 - the value in each frame is changed by the same amount. so, if you keep a note of what number was added to each "global gain", you can just subtract it again, and you're back to where you started.

mp3gain isn't truly "lossless" if you don't keep a note of the amount you changed the gain, because you can't get back to where you started. However, the process doesn't introduce any rounding errors because the global gain value can only be changed in discrete 1.5dB steps.


wavgain, or any type of normalisation applied to the .wav file, is lossy, because you're working with a fixed precision (e.g. 16-bits), and every change of gain is followed by a rounding process. To change the gain, you'd multiply every sample by the same number, and then you have to round the result. However, unless you're very lucky, you can't multiply by a number, round the result, divide by the same number, round the result, and get back to where you started. You'll get approximately the same number, but it won't be exact.


replay gain usually suggests that the audio file should be quieter than it is. If this is change is carried out within the decoder (or at least before clipping or truncating the decoded samples), then any clipping that would have been introduced at the decoding stage of a lossy codec will be eliminated. It can't do anything about clipping on the original CD - it can only prevent additional clipping due to the lossy codec (e.g. mp3).


Hope this helps,
(still recovering from your avatar, but it's very nice!)
Cheers,
David.

LMAO

thank for your answer. this is the first time i understood it. thank you very much (!)

I saved this one to my best of hydrogenaudio file on my hd.

just so that i understand this, replaygain for FLAC will not cause any rounding errors? all i want to do is take a few tracks off different CDs, and change the gain so that they all have the same perceived loudness, but without introducing any rounding errors like i get with, say, wavegain

FLAC, Monkey's Audio, Shorten, and other lossless audio formats are just that, lossless. No data is lost or altered beyond recorvery, unlike mp3, so you can always decompress a FLAC file and have the original unchanged wave again. Just like how ZIP files compress any file without any information being lost so that the original file can be regenerated. But these lossless audio formats are specially tailored for audio and the playback of audio and thus can compress and play music much better and more effetiently then ZIP files can.

ReplayGain tacks on data about the loudness of a song onto the FLAC file (just like how id3tags tack on song info onto mp3's). When the FLAC files are played back, the ReplayGain information is read, and the audio is normalized down and then played so that that the song is percieved to have the same volume as any other songs that have ReplayGain applied to them. And none of this alters the audio data in the FLAC file.

>>>'When the FLAC files are played back, the ReplayGain information is read, and the audio is normalized down and then played so that that the song is percieved to have the same volume as any other songs that have ReplayGain applied to them. And none of this alters the audio data in the FLAC file.'<<<

so what you are saying is, replaygain will not introduce rounding errors in FLAC files (just like mp3gain doesn't introduce any rounding errors in mp3 files), even when the FLAC file is decoded back to wave?

if this is so, how to decode the FLAC file back to wave after the FLAC file was "replaygained"? i'm using winamp 2.81 to play FLAC files. do i just use the nullsoft diskwriter plugin?

(sorry about the continuous wave of questions, i never used FLAC before )

Using the diskwriter is always a bad idea IMO. All those input plug-in settings and EQ and DSP, there's just too much potential to mess it up. Use a simple frontend like FlacDrop. Since FLAC's ReplayGain works with a few bytes of information at the beginning of a file (just like MPC, for instance), decoding without any options will give you the original WAV.

Yeah, but... using Winamp's diskwriter is the only way to to decode a FLAC file with ReplayGain info, which is way faster than decoding with flac.exe and then WaveGain'ing. I guess the main purpose one would want to decode FLAC files with ReplayGain info for would be burning an Audio CD - Winamp is perfectly OK for that IMHO, as long as all DSPs are deactivated. (There's not much to be misconfigured in the input plugin...)

Yes... but part of why this is not available is that Josh does not want to introduce any sort of "lossy" compromises, and allowing ReplayGained decoding of FLACs would mean that the output was not strictly the original WAV. That said, I still believe it would be a useful feature. For one thing, although one might be able to directly copy the ReplayGain information to a new file (assuming one is transcoding, and not decoding) there is still very little hardware support for ReplayGain. And if one is decoding FLACs to burn an audio CD, WAVs have no sort of ReplayGain parameter at all; WaveGaining is still an essential step.

john33, is it safe to ask how much work WaveGain needs? Between the earliest versions and the current seemingly stable release (0.9.6), the output changed dramatically. (That is, starting from the same original WAV -> WaveGain would produce slightly different results with many versions.) Are the calculations "frozen" now?

Also, I recently encountered a puzzling test situation: Low-bitrate (32kbps/22kHz) MP3s decoded with MAD - to a 16-bit WAV - do not seem to be analyzed by WaveGain, although the "INPUT FILES" description says it will accept 1 or 2 channel audio, at a whole range of sample rates. Is this a bug, or did I do something wrong?

WaveGain already possesses the ability to set bit-depth. Would it be possible to add a sample rate converter, for all-in-one processing?

One last request (and I've mentioned this one before; I know CoolEdit has a similar feature... but I use a different Wave editor and would prefer a freeware solution): Is there a way to add a "clip restoration" feature to WaveGain, so that albums which were deliberately mastered too loud - see anything on the Experience Hendrix label, and most other albums mastered by Sterling Sound Studios in New York - could be "repaired" for personal listening?

- M.

I was wondering if there was any tool similiar to wavegain that would allow me add something like replaygain but then bump the volume up so I would be using more of the headroom on a cdr. For instance, if I just wavegain and then burn, I end up with a very quiet audio cd.

ps, this is for use in my car.

this is precisely what i want to do. i want to make a compilation CD from several different CDs that i have. i want all the tracks on the compilation CD to be equally loud

until now, the only way to do this is to use wavegain, or normalise using a digital sound editor, both of which cause rounding errors in the wave file. but if i can decode FLAC files with the replaygain tag to .wav without any rounding errors, that will be fantastic

Any time you adjust the the 'gain' whether it is to the original wave, or to a FLAC file on decode using the repaygain tags, you will change the data. To describe it as 'lossy' is not really the correct description since people will confuse it with the lossy nature of a lossy codec, but the data is permanently altered. However, don't get hung up on this, the sound will be different because the volume has been changed but I do not believe there is anyone who can audibly detect where any information has been 'lost'.

If you want to level off the volume, but to a higher level, you can apply a 'manual gain' to the process, see the options. In this case, I would also recommend applying the 'hard limiter'.

Also, although it never really seems to be mentioned, as such, normalisation is every bit as much a lossy process as is WaveGain. By the same token, 'resampling', however good the algorithm may be, must also be described as a lossy process.

Any process that changes the original sample in any way whatsoever MUST be described as a lossy process. So if you want to apply any of these processes, do so and judge the result using your ears, not the theoretical argument!!

They're both the same since FLAC is lossless. It doesn't matter which one you choose. You'll lose a slight bit of resolution either way.

You are right, but as a general principle, when you *can* choose between a lossy and a lossless procedure, it makes most sense to use the least lossy one.

An example what I mean:

wav -> wavgain -> vorbis -> player

wav-> vorbis -> replaygain -> player

The first sequence quantizes to 16 bits twice, whereas the second one only does once. The second does the replaygain volume scaling basically in a lossless manner (well, at 32-bits, but you know what I mean), whereas with the first one you'll introduce additional quantization noise.

Given the choice, you should pick the second. Whether it's audible or not - probably not and certainly not easily, but given the choice, why take the risk?

This is the wrong thinking. Loud CDs are not using "more of the headroom", they are reducing the headroom almost to zip zero. This has several negative effects on audio quality.

If you can't turn up the radio in your car some more, but are in need of more "bump", i suggest using WaveGain with the following options: "--album --gain 3 --dither 3 --apply". This will adjust it to 92 dB. That's enough for every car radio. If you turn it up more, you'll have to live with the sideeffects (clipping, distortion...).

ok guys, now i'm really confused

can someone please confirm this:

1) mp3gain is regarded as "loseless" because you can change the gain of an mp3 file, and if you change your mind, you only need to put in the original gain value, and now the file will have the same CRC as the original mp3 file. is this correct? if this is correct, can you do the same thing with replaygain for FLAC?

i assume the replaygain process can be characterized as "loseless" as long as the FLAC file is not decoded to .wav. if the FLAC file is decoded to .wav with the replaygain tag, then obviously there will be loss of resolution, especially if the level is reduced by a significant number, like 15 dB

2) if you change the gain of a wave file, for example, the process is entirely different. once you change the gain of the original wave file, if you change your mind and want the original gain back, you can apply the original gain value except that the result will be different than the original wave and the CRCs won't match. am i correct?

the problem i keep experiencing are quantization errors at low levels. john33, this is not only based on theory, but on actual listening. i listen to a lot of my music with headphones, and these errors are in fact audible, especially during the tail of a fadeout. as the levels decrease during a fadeout, the sound gets more "distorted". i just want to avoid "re-dithering" whenever possible, if possible

If you APPLY ReplayGain, you change the sound.
It will even sometimes make clips sound better,
because makes clipping less audible.

If you add dither, problem you described becomes very much less audible.
That's why sound modified by *Gain should be output at higher resolution
(eg. 32bit for 16bit sound) and then dithered back to needed one.
This isn't possible with MP3Gain, because program doesn't know if the sound was modified or not.
Then it is highly unlikely you'll other difference than in volume, even in fades.
(Unless you'll modify this tag by hand to contain some weird value)

ReplayGain, contrary to WavGain, doesn't modify audio data in file, it is applied on playback.
Thus, it can be ignored and doesn't need to be compensated for.

<edit>
If you'd like to use it in portable, you have to apply it nonetheless...
Be warned that Winamp2 FLAC plugin AFAIK doesn't know about replaygaining.
Only Vorbis plugin supports it.
</edit>

Yes. BTW, lossless, not loseless.



Josh's concern with FLAC is a different one. FLAC, like MPC, works with some ReplayGain bytes at the start of a file, so the compressed audio data is not touched. However, when you decode to WAV with Replaygain adjustment, it can't be considered lossless anymore, with the rounding, dithering and whatnot involved. For the term "lossless", it's irrelevant if the changes are audible or not. Having said that, i can understand that it would make things easier for some people if the Replaygain info could be used for decoding to WAV, not just playback. However, a warning should be shown that the process isn't lossless anymore.




It has little to do with the amount it gets reduced by. You can multiply by 0.99 or 0.55, there will be rounding errors in both cases.




Yes. WaveGain processes the whole WAV file and directly applies dithering etc., since a WAV file, well, cannot be decoded any further, so to speak.

ahhhaaa, now i understand this a lot better. thank you lads

in the case of preparing files for a cd wouldnt there be nice if something like 'wavegain -y --adjust2maxpeak' would exist or am I missing something?

smok3 I believe volumax does that.

Uh... what would maximizing be good for? High volume does not correlate with quality.

uhmm, it would correlate with (at least theoreticaly) higher dynamics, no? (i mean, the scaling should be as small as possible, since most of the original cd's are maxed this days), i mean, in relation to each other and track type of replaygain of course (like adjusting the gain to max peak of the N files).

you sure?
http://www.prossalonsupply.com/en-us/dept_94.html

Haha, thats not the volumax I'm talking about. I'm talking about the volumax.zip from John33's mirror 1. Go to programs, others, and download the volumax program. Here are the posts about volumax here on HA.

http://www.hydrogenaudio.org/forums/index....2656&hl=volumax

http://www.hydrogenaudio.org/forums/index....2368&hl=volumax

Sure does!

hmm, what command should i use? (this does not look right, is there track replaygain integrated?)

If you check the options, you can do do it on a track or album basis, just like WaveGain.

still looks like a peak normalizing 2 me, i must be doing something wrong?

It's like we speak a different language.. There are no "higher dynamics". You are effectively reducing the dynamic range. Many DACs have serious problems already below full scale, with distortions starting at -3 dB for instance.

what exactly is your suggestion for prepairing the audio cd made from different sources when i want some sort of equal loudness and i only need the tracks to sound equal on that cd? (I understand that max peak is actualy not a good solution, but that peak would probably happen to only some of the tracks when using RG, no? (sounds better than applying some imaginary bonus and then limiter imho, the bonus level should be calculated from the highest peak of all the N files... and iam willing to adjust that volume knob for each new cd of course )

WaveGain in Radio mode, with +3dB on the reference level and dithering.

Commandline: "--radio --log --gain 3 --dither 3 --apply" (-r -l -g 3 -d 3 -y). IMO a hard limiter isn't needed. Works good on my car stereo.

i still got a question:

you say that it is NOT LOSSLESS to apply gain to a wav file.
and you say that it is LOSSLESS to apply gain to a flac file for example.

What is lossless? the result i hear or just the fact that i can decompress the flac file back to wave without the applied gain ?

in case the flac file with applied gain is lossless compared to the original wav file (when listening to it) i would get lossless wav files with gain when using the diskwriter plugin when playing flac files with the applied gain.


i think it's like that

LOSSLESS (PLAYBACK): wave file (NO GAIN); flac file (NO GAIN)
NOT LOSSLESS (PLAYBACK): wave file (GAIN); flac file (GAIN)

LOSSLESS (STORING): wave file (NO GAIN); flac file (NO GAIN), flac file (GAIN)
NOT LOSSLESS (STORING): wave file (GAIN)


???

If you apply WaveGain to a wave file, the digital representation of the samples is changed. NOT LOSSLESS.

Apply ReplayGain to FLAC file. ReplayGain values stored as metadata, samples unchanged. LOSSLESS. BUT, playback using replagain values to adjust volume is NOT LOSSLESS, samples are changed on decode for playback. Decode to file using replaygain values is NOT LOSSLESS. Decode to file ignoring replaygain values IS LOSSLESS.

Does that make it clearer?

Basically, anytime you apply ANY process that changes how the samples were when first ripped, that process is NOT LOSSLESS. Normalise, re-sample, WaveGain, downmix are all LOSSY processes.

then i understood it right.
should i anyway use replaygain (vorbis) ?

If you plan to make use of it for playback, yes. Otherwise, there's no real point!!

Damn, John beat me to it for Xenion's question (hmm, I hadn't realised it had been over an hour since I refreshed this thread ). Anyway, here goes...

---------------------------------->8----------------------------------

Xenion:



No. The reason why replaygain'ing FLACs, Oggs, MPC etc. is considered "lossless" is because the ReplayGain data is written to a tag, leaving the original sound data unchanged. Once you actually make use of the ReplayGain changes, of course the sound will be (slightly) changed because of rounding errors resulting from the volume change (on playback and when decoding to WAV), so the actual usage of ReplayGain is not totally lossless, it's just the storage of RG information that is completely lossless.



smok3:

Wait a minute - you are trying to do the following, right? You'd like all tracks on the CD to have equal volume, but the level of the entire CD raised as far as possible? (You don't want each track maxed out, but the CD as a whole?)

If that's the case, you need to do the following:

- Decode replaygain'ed files to WAV in track mode, or apply WaveGain in track mode
- Apply VoluMax in album mode. This will raise the volume of all tracks (applying the same change to every track) until the first sample hits 100% (or whatever you specify with the -l option).

CU

Dominic

@volcano. thats the way i thought it has to be...
btw I vorbisgained why worry from the brothers in arms album: +12db :-) lol

IIRC, that's the only album that I've ever processed thru WaveGain in Album mode that recommended no change!!

lol maybe you've got the remastered version. the only dire straits albums i have which are not remastered are brother's in arms and communiquè. there is even a user's manual in it how to handle cds (3pages). i guess it's from ~1990 when cds came out.

Actually, it dates from 1985 when it was first released and was pressed in (West) Germany!!!

mine too !! :-)

Made in W. Germany by Polygram
© 1985 Phonogram Ltd London

but i don't know if it is the date when my cd was pressed or the date when this is was released. because it came out 14th May, 1985, so this may also be the release date and not the pressing date

I don't know when this cd was baught because i actually stole it from my dad some years ago

Well I bought my copy in about 1986!! It was around the time I first got a CD player.

tnx, just out of couriousity, where did you get the --gain 3 ?



ic, i thought there is one-step process possible with volumax, cos that is lossy+lossy now
(and yes, that was the idea)

When compressing a whole cd(several different tracks) using FLAC for Replaygain, should I assign the value radiogain or ablumgain? And for one hour long set(for techno music), which should I use?


-Darin

:alien:

If you want to preserve the volume relationship between the tracks in the album, use album gain. Otherwise, use track gain.

When calculating the gain, always use Album Gain. This stores both track and album values instead of only the track values.

When you have both values in the tag you can choose on the audio player which type of gain you want to use at time of playback.

Track Gain may be preferred for parties and shuffle mode.

Album Gain is almost as good for those and is also best for continuous listening to a whole album.