Is there any chance of doing something like a lossless mp3? Iv seen it on other sites where u encode to mp3, calculate the difference, and store the difference elsewhere. Would this be possilble with mp3? (id like to do it, because my MP3 player ONLY reads MP3) so that, id get the encoded mp3 on my player, but if i needed to, could get the original, (or close to it) back again.

It's possible...but I don't think there's any program that does it automatically.

ok, so is there any chance of doing this manually?

I think you're much better off just encoding to flac, ape or other lossless codec, and then transcoding the files you want to use on your portable to mp3. I believe what you described is theoretically possible, but there's no easy way to do it.

- Matt

I agree. Store them in lossless (APE or FLAC). Storing the difference file otherwise will not cost less in space and will not be playable (will be useless without the mp3 files actually). You said you want to store the difference elsewhere, so you'd be better off storing the lossless encode there which wouldn't incur any additional space cost then the difference.

I have already done this, but id prefer it, if ALL of my audio was on the portable player and none left on the PC. Space is not really too much of a problem as it is 20GB, but i thoght it might be a good idea to do this. Isn't there any way to do it, complicated or otherwise?

Problem is also that you need the exact same decoder you used when creating the difference file.

That isnt a problem, i could store the decoder on the portable player along with the audio. How do you create an MP3 Difference file anyway?

This would require specially modified encoder,
which needs to save thrown-out bits to the file. (possibly Huffman-compressed too)

There's no such thing AFAIK, but since LAME source code is free, anyone willing can add it.

isnt there any way this could be done in CoolEdit or something?

This whole thing will only make sense if the losslessly encoded difference file is significantly smaller than the losslessly encoded original file

no, because lossless files cannot be played on my Mp3 player, only Mp3's so, if i could store the Mp3 and the difference on, it wouldnt matter. I could still make the original file, yet still listen to the mp3 too.

ok, I think i may have a way using cool edit 2000. If, you open the mp3, and the Wav, then (delete all silence from both) copy the mp3, mix paste with the wav, but press Invert. This then leaves u with the difference which u can save for later use. The only thing is using this method, clipping seems to occur. Can anyone think of another way to do this? or to stop the clipping?

If you can store an mp3 and it's difference file on the player.. why can't you just store an mp3 and a FLAC lossless file on the player? You could still play the mp3, and you would save a lot of time just encoding a wave to flac than having to decode and subtract the mp3 content from the wave.

From what I've seen, you can set EAC to compress to both FLAC & mp3 at the same time (if you are extracting music from cd's).. so that wouldn't take you any extra time.

The difference file, uncompressed, is exactly as big as the original file, uncompressed. The difference file, compressed may or may not be somewhat smaller than the original file, compressed. It may even be larger if the noise introduced by the lossy compressor is big enough to throw off the lossless compressor. So, store both mp3s and flacs on your portable, it's easier.

yeh, ur rite, i just tried it. It does work, but when losslessly compressed, its bigger than the orginal compressed. Oh well, bad idea. Sorry

Yeah, that was my original point But I guess with WavPack lossy the difference file is smaller. I don't know how it manages that...

In WavPack hybrid mode, Lossy file + correction file is 1-2% bigger than a lossless encode.

ok, so it is a bad idea, but try listening to the difference file! u can tell just how much mp3 cuts out. Never knew it was that much!

Scary, isn't it? Check this thread...

Lossless Mp3 doesn't exist i think but, dont quote me on this! Anyway, --alt-preset standard with 3.90.3 is transparent and sounds the same as the original cd (IMO).
EDIT: i changed the font and wording

A ) Don't use big/bold fonts if there isn't really another way say it.
B ) To say something *that you think*, you'd better ask first, and talk in case of correctness.
C ) MP3 (except maybe freeformat) can *NEVER* be transparent IN ALL CASES, so telling such a thing is useless. I know of samples that not even in -api are transparent.

man he talks crap dont he. does he think i dont know about -aps? dumb or wot

If you want lossless mp3 encoding then you won't get it! Don't you get it? mp3 wasn't made to give you lossless encoding.

And if you claim that you can hear those a difference between a well encoded mp3 and the original then you are lieing!

Music I encoded with lame --r3mix preset are most of the time under 160kbit average and CRYSTAL CLEAR!

I'd suggest to you that you just encode your music with a state-of-the-art mp3 encoder with tweaked settings (cutting off the frequency at 18khz e.g. har har har) and store your lossless encoded dupes (e.g. using ape) on CD-Rs...

Have you bothered to read the forum rules? Have you bothered to do any research whatsoever before posting here? Have you taken the time to perform any ABX testing at all of the settings you advocate?

Most of the people here, myself included, are able to tell the difference between an r3mix encoded MP3 file with confidence using double-blind ABX testing. Trust me, I have much better things to do that "lie to people" on online forums.

Saying that <160kbps r3mix MP3 files are "CRYSTAL CLEAR" is an absolute untruth -- a total fallacy -- and continued statements of this sort will do nothing but get you banned from this commuity. Take the time to read the rules or don't bother posting.


Your wasting my time and the time of everyone here. Har har har is right.


Stop using bold and large fonts to post regarding something you admit to not knowing about. Your wasting my screen space. Anyway, I doubt you want to draw any more attention to your worthless post than it normally would. It's not even like your post was funny like Fandango's post is to anyone with a clue. =)

Anyway alt-preset standard encoded MP3 filea are usually transparent -- but not always so. That's why the extreme and insane profiles exist, and even these presets have samples that require more bits than any standard complaint MP3 format can provide. Some of samples recently discovered have passed ABX testing using ~500kbps MusePack.

Total transparency in MP3 files in all circumstances is impossible. Using alt-preset standard is the best you can come to transparency on most content. Using higher settings for "safety" is likely just to waste space. Tuning of the psychoacoustic model will help alot more than throwing more bits at your files. Most of the samples that produce the horrible artifacts aren't "fixed" by using extreme or insane.

The closest you may come to lossless in MP3 is a 640kbps freeformat MP3 file produced by a recent version of LAME. It will most likely not work on your consumer grade MP3 player decoder.

Which version specifically? And do you have a link settiing up the encoder?

I'd like to try and test this on my disc based player. At worse I'm only out a disc.

These freeformat MP3s aren't playable by most players, even software ones.
Additionally, there are some limitations:

FreeAmp up to 440 kbps
l3dec up to 310 kbps
mpglib up to 560 kbps
MAD up to 640 kbps

So, you'll most likely waste a CD. Try it with a CD-RW first.

Just in case: Have you taken into account that mp3 encoding->decoding adds samples in the beginning (and in the end) resulting in an offset? For correct results this offset needs to be cut away.

Listening to the differences tell you nearly nothing about codec's quality.
Codec equipped with good psychoacoustic model will cut much, but it wouldn't be audible.

As you could have said this in your post before I guess this is a reply to my post. Of course you're right - I just wanted to point out that no matter how useful these differences are, you have to take offsets into account to get correct results. (E.g. if the wave substraction is done correctly the result won't contain clipping in most cases.)

You're right. Maybe I'm just being slightly paranoid lately.