Some more question to you, people. I believe you can remove my doubts.

When I get some mp3 music in 320 (and sometimes in 256) kbps CBR, I'm used to recompress those mp3 using LAME VBR (-V2 --vbr-new for now), because otherwise filesize stays too large.

So, the question is: IS IT SAFE to recompress mp3 to another mp3? I know that transcoding between two lossy formats (e.g. OGG→MP3) kills quality. But here is mp3→mp3. However that may be, there are different mp3 codecs, so may such a recompression cause futher quality loss?

Transcoding between any two lossy formats (mp3 to mp3, aac to mp3, mp3 to ogg, etc.) will always result in a quality loss. So, yes, when you take a 320kbps mp3 (or 256kbps mp3) and re-compress to a Lame -V 2 mp3 (I don't know why you would transcode the 256kbps mp3), that resulting Lame mp3 will be of less quality.

However, the big question is: Do you notice the decrease in quality? If you don't notice the decrease in quality then don't worry about it. You can always setup a blind ABX test using foobar2000.

I think I won't notice. But will the following way also cause loss?

1) taking LAME 320 kbps
2) decoding to wav with LAME
3) encoding back to mp3 using LAME VBR2 new

…Really I think that will cause qualty loss however (theoretically). Because even 320→wav→320 will (even if using the same codec).

look at it this way: when you compress an original track to an mp3 file, you shave off information to reduce the size. that information can not be restored. so now, with what you're trying to do, is shave off even more information off an already shaved mp3 file. this will result an even bigger quality loss than if you compress the original track to mp3. you see where this is going? this applies to all lossy formats, image or sound.

Yes, I meant the same... It is something like "100 cycles of mp3->wav->mp3", but much better because there are only 2 cycles (or even 1, if cycle consists of encoding and decoding operation ) Am I right?

mp3 -> anything_else implies mp3 -> wav -> anything_else ( compressed data needs to be decompressed in order to be recompressed )

The reduction of quality from mp3 -> mp3 as opposed to mp3 -> ogg, for example, might even be higher, especially if using the same encoder, because it will be adding the same (or very similar) distortions, increasing them.

That wouldn't cause any differences, decoding to .wav can't simply restore the data the MP3 encoder removed during the first archiving process. But due to the high bitrates of both your input and output files I'm very sure that you won't be able to distinguish them from each other. The transcoding loss we're talking about in this topic might be ABXable if you decided to create a bunch of, let's say, -V5 files out of a -V2 collection. But transcoding from the insane preset to -V2 should be safe in my eyes. If you're uncertain about this theory, simply transcode a few samples and ABX them.

Look at this test. The result says it clearly:

Recompressing from one lossy format to another one causes dramatical loss of quality. Recompressing from MP3 to MP3 is the worst case.

It's a non-representative personal test of Guruboolez. To get more reliable results you should carry out another test with a much larger audience, including some non-audiophiles who haven't trained their hearing for ABXing artifacts over the years.

Hm… it is strange for me that recompressing mp3 to mp3 is the worst case, but I’ll have this in view, thank you.

BTW, I think results of recompressing 320 to VBR2 should be better than the results of recompressing 128 ABR. Much better.

Instead of transcoding, why don't yo have a look on mp3repacker? It is loseless and in some cases (for example with 320 kbps source files) gives really good results.
~

Ok: 320kbps --> wave --> V2 is not a good job for audio quality.

My question:

could be freeformat --> wave --> V2 a better job ?

(Sure i know lossless --> wave --> V2 is the best job ! , but this is not the right answer to my question )

The difference in sound quality between a freeformat file and a CBR 320 file is probably undetectable, so really you are asking if there is any difference between the same thing. I can't see how whether the initial MP3 is 320 Kbps or 640 Kbps is going to make any difference, you are still going from lossy to lossy which is never ideal.

640kbps free format will give very different results from 320kbps. The coding noise is much lower in amplitude, and of a completely different distribution. Or it was when I last used freeformat, which was several lame versions ago.

Transcoding is covered well in the FAQ.

Cheers,
David.

However with higher bitrate there is less quantization noise. Also one could imagining disabling the Psycoaccustic model, which would benefit transcoding (like transcoding from Wavpack Hybrid).

But why would you bother when a lossless codec averages (in my experience) around 850 Kbps? An extra 25% bitrate and you get lossless.

Transcoding free format files seems to be a lot of bother for not much gain.

Yes !

This is the answer i gave myself !

Freeformat is the same as CBR320kbps for my ears, but it's not for the Lame encoder !!!

It's not the sound quality at that stage which is important, it's the difference in data which has an effect on the audible qualities of the second compression.

Sure, but why even bother transcoding from free format files? The bitrate difference becomes so small that you may as well make lossless files anyway. Lot's of players can play lossless files, but how many can play free format MP3s?