Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: lossless mp3 - mp3HD (Read 98233 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

lossless mp3 - mp3HD

Reply #50
http://www.hydrogenaudio.org/forums/index....st&p=544529

Quote
That's why SLS also operates in the frequency domain which is incompatible to the MP3 filterbank.



I actually didn't test it.  So, this was me guessing.  The argument is that MP3 uses two filterbanks that are concatenated (PQMF + MDCT + alias reduction) and SLS is based on the MDCT only.  But maybe this concatenation is "similar enough" to what a plain MDCT would produce given a good window function.  But maybe they really use an integer MP3 decoder and update the time domain data using the "HD" data.  I don't know.


It's not impossible to make a lossless PQMF, DTS MA has this for example. Add lossless IntMDCT and lossless alias reduction (possible because it is only some butterflies) and you have a lossless mp3 filterbank.

lossless mp3 - mp3HD

Reply #51
It's not impossible to make a lossless PQMF, DTS MA has this for example. Add lossless IntMDCT and lossless alias reduction (possible because it is only some butterflies) and you have a lossless mp3 filterbank.

Hmm... sounds like there would be a lot of intermediate steps/roundings involved. The beauty of the MDCT is that you can factor it into a couple of butterflies and a type-IV DCT. This kind of DCT is an involution which makes "multidimensional lifting" applicable. The "multidimensional lifting" approach for a reversible integer-DCT is one of those that approximates the real-valued DCT pretty well. I don't yet see how the mp3 version could be as good w.r.t. approximating the real-valued filterbank.

The alternative would be using a deterministic integer mp3 decoder + correction in time where the spectral shape of the "correction" noise could be infered by the mp3 frame's side information (scale factors, code books).

Cheers!
SG

lossless mp3 - mp3HD

Reply #52
As concept this codec has great potential (maybe can bring lossless mainstream)

But it fails in practice. That hack is just facepalm.

lossless mp3 - mp3HD

Reply #53
As concept this codec has great potential (maybe can bring lossless mainstream)

But it fails in practice. That hack is just facepalm.

Quite honestly, I don't think that lossless will ever be mainstream.  Looking at listening tests lossy codecs seem nearly transparent to people at 128kbps and iTunes and Amazon mp3 are using 256kbps and I don't here complaints about audio quality.

lossless mp3 - mp3HD

Reply #54
I WAS all excited when I read about it on some website. I presumed they somehow made MP3 lossless, (as in I could encode stuff that would play in full quality on any MP3 player). Sigh. I just wish Apple and and Sony would support FLAC!

lossless mp3 - mp3HD

Reply #55
Well... I don't get it... I extracted a correction file from a 190kbps and a 130kbps VBR encoded file (done with foobar factory settings and LAME 3.98.2) containing all the differences between the mp3 and the original wav and compressed that corrrection file with FLAC and the filesizes overall are almost on par:

Test track: AC/DC Smash N Grab
wav filesize: 41.4MB
complete flac file (lossless): 34.1MB
mp3 (190kbps VBR) + flac'ed correction file: 6.2MB + 29.6MB = 35.8MB
mp3 (130kbps VBR) + flac'ed correction file: 3.9MB + 31.7MB = 35.6MB

Test track: Orchestral track
wav filesize: 17.9MB
complete flac file (lossless): 11.3MB
mp3 (190kbps VBR) + flac'ed correction file: 2.6MB + 9.5MB = 12.1MB
mp3 (130kbps VBR) + flac'ed correction file: 1.8MB + 10.1MB = 11.9MB

Tho I had to correct those aligning issues described above manually (cropping a selection of the mp3 which was longer than the orignal wav both in the beginning and at the end... maybe using VBR caused this?). At least that way (if properly implemented) you'd be able to get back a lossless file and still have everything on mp3 as well without major sacrifices and spare the overhead of full flac plus mp3 files. Just a player that can stream both at once would have to be implemented as well and you'd be all set I guess.

lossless mp3 - mp3HD

Reply #56
Unless they invented a fully lossless MP3 format, that could be decoded by any MP3 player, then this new format is useless. FLAC all the way!

lossless mp3 - mp3HD

Reply #57
While I agree with everyone who says that this is a completely useless format, I wonder if a proper lossless encoder which used FFT analysis and frequency domain representation for predictor might achieve more easily compressible residuals...

lossless mp3 - mp3HD

Reply #58
Quote
I wonder if a proper lossless encoder which used FFT analysis and frequency domain representation for predictor might achieve more easily compressible residuals...


white noise cannot be compressed at all losslessly.  you can make an audio file that cannot be compressed to even half size by any compressor simply by mixing a loud track with half-scale noise.  what you describe is not a usable metric for evaluating compression because it is too dependent on the input.

state of the art lossless compression is close enough to the noise level of the input already.


lossless mp3 - mp3HD

Reply #59
Quote
white noise cannot be compressed at all losslessly. you can make an audio file that cannot be compressed to even half size by any compressor simply by mixing a loud track with half-scale noise.

Of course white noise won't compress well losslessly when it's basically completely random data! However, listening to music we can all very easily hear that what we hear is not random - or at least not even remotely as random as pure white noise. That's the reason why lossless audio compression works at all, actually.

As for whether or not better lossless audio compression than what's possible today can be achieved... I think it's possible. Look at for how long zip was the best general compressor available, and then rar came and blew it away. And then, 7zip came and blew both of them away. I think there's still room for improvement. As for how much more will be possible... that I really can't guess. If you forced me to estimate, I'd say (with great uncertainty) that at least 5% - 10% more is possible. But that's really a shot in the dark.

lossless mp3 - mp3HD

Reply #60
frequency domain prediction has been tried.  before lpac there was ltac.  there are several older threads here about this.

lossless mp3 - mp3HD

Reply #61
This format is just as useless as Pulseaudio.
"I never thought I'd see this much candy in one mission!"

 

lossless mp3 - mp3HD

Reply #62
frequency domain prediction has been tried.  before lpac there was ltac.  there are several older threads here about this.

Oh. Well I'm sort of new here so I apologise for not knowing that there already was a discussion about this.

lossless mp3 - mp3HD

Reply #63
I want to know if exists a plugin for foobar2000 for decoding mp3hd?



lossless mp3 - mp3HD

Reply #66
Is this encoder still available somewhere out there or?

Have been trying 2 get my hands on it but the all4mp3 website is down.

lossless mp3 - mp3HD

Reply #67
The web site of the German IT magazine CHIP has apparently mirrored the encoder package to their servers. Haven't tried it though.

http://www.chip.de/downloads/mp3HD-Toolkit_13001528.html (click on "Zum Download")

Chris
If I don't reply to your reply, it means I agree with you.


lossless mp3 - mp3HD

Reply #69
A label I am friendly with just started selling MP3HD downloads. 

I told them to stop that right now.

In the meantime, when I use XLD to convert to FLAC, is it using the full lossless component from the MP3HD files? I can't find a way to tell.

lossless mp3 - mp3HD

Reply #70
And... why are people still using or ever used mp3HD??

lossless mp3 - mp3HD

Reply #71
Because it's "HD".