Ok, i have sort of talked about this subject before. What id like to acehive is kinda what OptimFrog dualstream does. U can make an encoded file, and save the correction file too. Decoding both the encoded file, and correction file together, makes the original file.
I have tried this with MP3, but its not effective. I used it by encoding the file, then finding the difference in CoolEdit, but the thing is, if u save this difference file, it is the same size as the original was anyway!

Anyone know of a way of just saving the difference in an effective way?

I don't believe it makes sense to open the same discussion after a week or so. To answer your Q, just use your favourite lossless codec to archive your original wav file.

I don't remeber if you explained this the last time, but why do you have to do it this way?

Edit: It seems like you only make thing difficult for yourself and less flexible. Painting yourself into a corner IMHO.

didn't we answer this before: you losslessly compress the difference file using FLAC or Monkey's Audio or whatever.

you'll find that the losslessly compressed version of the difference file is similar in size to a losslessly compressed version of the original.

In other words, there is no practical use! If you want an mp3 version and a lossless version, then make two separate, independent versions!

Cheers,
David.

lol sorry, i dont think u all quite understand. Im not saying the same thing as my last posts. I kno it all seems quite stupid to you, but my MP3 Player, can ONLY play mp3's, nothing else, not even wavs. Now, i wanted a way, that I could listen to the MP3 on the player, but if i needed to (like broke the cd or woteva) get the original files back. That is the only reason. I have the files encoded in ape already on my mp3 player, but i cant play them. Just wondered if I could save myself space by doing this.

As I understand, this is exactly the same question as the last time.

If you want to reconstruct the original you will have to store the difference file losslessly, right? And you will have to use the same MP3 decoder you used when you did produce the difference file.
This difference file will be about the same size as the losslessly encoded original, you won't win anything.
It will be much easier to reencode an MP3 from the original (compressed) WAV than first having to calculate the difference and then, for reconstruction, adding decoded MP3 and difference values.

Well, now you know that you can't save much (any?) space doing this.

Sometimes mp3 + recovery ape takes more space than mp3 + normal ape!

Cheers,
David.

i understand when u say that yeah, but look at optimfrog, that can make a correction file, which is much smaller than the losslessly encoded original

Optimfrog and Wavpacks lossy files are non-perceptual (they don't use a psychaccoustic model), so always the same parts are left out.

ok. Do i get the feelin im not liked here?

jim: not yet disliked, but you do sound like trolling.

MP3 is not designed for hybrid storage. So there is no easy way to achieve that. Try it yourself: compress the difference losslessly and compare it to the compressed original. Sometimes it will be smaller, sometimes bigger, the same on average. And it's less reliable: you have two files to worry about.

So you get the same consistent answer: you want to create two independent files.

What's unclear?

That works because the lossy and correction file are two halves of the same thing, which (with a little redundency and extra info) add up to the lossless file.

But an mp3 file is not half of anything - it's a perceptual coding that is not designed to be corrected.


In the two programs you mention, the split between lossy and lossless is largely mathematical, or statistical. Let me simplify it. Lossless codecs are mainly prediction coders - they try to predict what the signal will be in the future. Obviously the prediction isn't perfect, and you have to store more information to correct it - to make it lossless!

In hybrid mode, the lossy file is the prediction part - it contains the information to drive the predictor to get very close to the original signal. The correction file is the part that's left over - it's unpredictable - it actually has a very small numerical effect on the audio samples within the file, but it takes a lot of space to store because the correction data has to be expressed directly - there isn't a more compact way of expressing it.

(continued in next post...)

Contrast that with mp3. It uses psychoacoustics to remove data. The resulting file is mathematically very different from the original. The difference in amplitude on a sample by sample basis between the original and the mp3 can be huge - almost as big as the sample values in the orginal file!

What's why a correction file to make the process lossless can be almost as big as the original file - bigger sometimes!

In essence, encoding to mp3 first doesn't help with lossless encoding - it makes it harder!


Maybe you are a troll - but that's unlikely. What your'e suggesting does, at first sight, appear quite logical, and I thought your deserved an explanation as to why it doesn't work.

Cheers,
David.

thankyou, at last someone sees my view