New Lossless MP3 Format Explained 346
CNETNate writes "Thomson, the company that licenses the MP3 patent, has released a new lossless MP3 format called mp3HD. It utilises both lossless and lossy audio contained inside a single .mp3 file, and the files will play on all existing MP3 players. The idea is simple: lossless files on your desktop that can be transferred without conversion to iPods and MP3 players. The issue, it transpires, is that although the full lossless/lossy hybrid MP3 file is transferred to players, only the lossy element can be played back. A command line encoder can be found on Thomson's Web site."
The obvious problem (Score:5, Informative)
that you probably thought of when you read the summary ("So now I get a larger-than-FLAC sized file on my portable player so I can get 128kbps?") is acknowledged in TFA.
Complete waste of time (Score:4, Informative)
Relevant hydrogenaudio thread: http://www.hydrogenaudio.org/forums/index.php?s=55b656dc8cdb3b97da794e936b2a9b1d&showtopic=70548 [hydrogenaudio.org]
In summary, it seems like a fairly useless and poorly thought out format. To be clear, this WILL NOT play losslessly in a standard mp3 player, you must use a special decoder to get the lossless bit. It will only play the lossy component in a normal mp3 player.
Lossless information stored in id3v2 tags? Bad hack that will break just about every tagging program out there. File sizes much larger than real lossless codecs and encoding/decoding speed is much slower than flac. Also you can't have tracks longer than about an hour due to id3v2 size limits. Additionally, a full size flac file and 256kbit mp3 often comes in at a SMALLER size than this one monolithic hacked up mp3.
Nothing to see here people, this is a waste of time. Something like lossy/lossless wavpack hybrid is a much better solution.
Sam
Re:The obvious problem (Score:2, Informative)
Learn with history or make the same mistakes. (Score:5, Informative)
I dare say that this insistence on backward compatibility is going to kill this format.
If anyone still remembers, many years ago Thomson released the mp3PRO format.
It was a low bitrate MP3 with some added spectral band data that could recreate the original
music sound quality. So in theory, you could have the same quality for half the bitrate/size.
To my decaying ears, it sounded really good at the time... if played on the supported players.
But when you played these files in any unsupported player, which happened to be all of them
except for the Thomson's Player or the Thomson's Winamp Plugin, you ended up listening to
a HORRIBLE low bitrate sound quality, since the extra mp3PRO information was ignored.
And even worse: you had no way of telling if a file being downloaded was an original mp3 file .mp3pro or something like that, the mp3PRO format might have had some chance...
or a new mp3PRO file, since they both used the same file extension. Maybe if they had renamed
the extension to
Years pass... and now they are doing the same thing again.
Instead of focusing on a lossless mp3 codec for a specific kind of market/enthusiast, they are .mp3hd or something similar.
insisting in keeping backward compatibility with players using the same method as mp3PRO did.
And once more the files are going to have the same extension as the original ones, instead
of
I hope I am wrong, but this surely spells doom to me.
Re:The obvious problem (Score:3, Informative)
Re:This is useless. (Score:3, Informative)
What the hell are you talking about here? It might fail to allocate a 256 MB block if the machine doesn't have enough memory, or if the program decoding the module is running in the kernel and using kmalloc, but for the most part, applications do not have to worry about memory fragmentation. Virtual memory takes care of fragmentation for you, as only 4KB pages need to be contiguous.
The only time this wouldn't work is if the application that you're running doesn't have 256MB of its address space free. Unless the application is using close to the 2GB or 4GB address space the application is given this shouldn't be an issue.
Phil
Re:Why? (Score:1, Informative)
"real-time" in your case means that it can transcode the song while playing the audio. A four-minute song takes four minutes to transcode.
But the kind of real-time transcoding that would obviate the extra data (which would increase the "lossless" files by what, 10%? big deal.) is the transfer speed of the storage media. If it takes four minutes to transcode a four minute song, that's orders of magnitude too slow. A four minute song needs to take less than a second to transcode.
Re:Why? (Score:2, Informative)
> My P-133 could do better than real time encoding of .wav -> .mp3
That's odd, since l3enc on my P133 ran at a very small fraction of real-time. Heck, it took 1/4 of the cpu just to do real-time playback.
Evergreening (Score:4, Informative)
Sounds like evergreening to me
http://en.wikipedia.org/wiki/Evergreening [wikipedia.org]
Re:why? (Score:4, Informative)
In case it isn't bleeding obvious (apparently it isn't): The key to good compression is prediction. If you can predict the signal to within a small margin of error, then you only need to encode a small error correction stream. In this case, the MP3 signal serves as the prediction and the remaining data is the correction stream. This concept requires that the prediction is stable, and since the prediction isn't an algorithm but based on actual data, that data has to be delivered with the correction stream. So this isn't so much MP3 with additional information as it's a lossless format which happens to use an MP3 stream as a component and is formatted such that MP3 players recognize just that stream.
Re:why? (Score:1, Informative)
That is already possible using Vvorbis + Flac in a ogg container, the only problem is the player support.
Re:why? (Score:5, Informative)
The only use case can see is if you own a mp3 player with large storage that doesn't support playback of a proper lossless format.
With this you can keep and listen to the files on your mp3 player while also being able to decode them losslessly when you plug that player into a computer.
also given the filesize stats in the article it appears they aren't just bundling together a lossy and lossless format but actually making the lossless format build on the lossy format (either that or they have a lossless format that is considerablly better than flac).
Re:why? (Score:5, Informative)
I've seen some comparisons at another site [videohelp.com]. A 41 MB wave file gives a 20 MB FLAC, and 22 MB MP3HD. So if the MP3 was indeed a skeleton of the lossless portion, it isn't very efficient. It's the same size as a normal lossless format + a separate MP3, stuffed into the same file. Actually, I doubt the MP3 has any use at all in the lossless playback, but I am ready to be corrected if anyone can cite something and not just speculate.
Re:why? (Score:3, Informative)
But this new format makes two copies of everything, it only packages it in one file! It's the same thing as picking a mp3 and "attaching" a flac file at the end. The space occupied by the too is the same, but in only one file. If you had the two, at least you could save space in your portable players.
Btw, http://mp3fs.sourceforge.net/ [sourceforge.net] is great: I keep my flac dirs in music/flac and I mount them using this in music/mp3. The mp3 dir show me all the tracks in mp3, so I can copy them directly to my player, but in reality they're converted on-the-fly as the copy occurs, so the used space isn't duplicated.
Re:A Far Less Brain-Damaged Solution (for Linux) (Score:3, Informative)
Good call. I was about to suggest this myself.
I have used MP3FS and it worked perfectly.
It's *the* ideal solution for people like me who like to have high quality audio on their computer but are limited to MP3 on their MP3 player.
Re:A Far Less Brain-Damaged Solution (for Linux) (Score:3, Informative)