Skip to main content

Topic: mp3+compensation file=lossless (Read 6603 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • tfouto
  • [*]
mp3+compensation file=lossless
Hi all,

I think wavpack would be much more know and widespread if there was a possibility of making a hybrid mode mp3+"correction file", perhaps this idea would be more convenient or in place in the Lame development...

Neverthless i will keep using wavpack as my main archive format, no matter what.

Regards,

tfouto

  • Bourne
  • [*][*][*][*][*]
  • Banned
mp3+compensation file=lossless
Reply #1
I think that can't be done due to the psymodels they are based on.

  • pdq
  • [*][*][*][*][*]
mp3+compensation file=lossless
Reply #2
The reason that wavpack hybrid works is that its lossy file is a very close approximation to the waveform of the original, so the difference between the original and the lossy is a bunch of very small values that compress very well.

Lossy codecs like mp3, on the other hand, don't necessarily produce something whose waveform is close to the original, only that it sounds close to the original. The difference between the waveforms can be quite large, and therefore not very compressible.

  • SebastianG
  • [*][*][*][*][*]
  • Developer
mp3+compensation file=lossless
Reply #3
I think wavpack would be much more know and widespread if there was a possibility of making a hybrid mode mp3+"correction file", perhaps this idea would be more convenient or in place in the Lame development...

MPEG4 audio now includes the possibility to create files that contain two tracks: one AAC track and one SLS (scalable lossless) track which allows bit-perfect reconstruction using both of'em. This has been mentioned here a couple times already. For example: See the HD-AAC thread.

The way the "SLS correction track" works doesn't carry over to MP3, though. For reasons already mentioned coding the "difference signal" as time domain signal is not a very clever idea in case of base codecs like MP3 and AAC. That's why SLS also operates in the frequency domain which is incompatible to the MP3 filterbank. So, the only noteworthy approach that comes close to your suggestion is the AAC+SLS combination.

Cheers!
SG