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: ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones (Read 8137 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Hi,

I ran some tests with different softwares and found that the replaygain values may significantly differ.

Here are the results I got in the following format
  OS / software : track gain, track peak, album gain, album peak

Code: [Select]
Al Di Meola - Elegant Gypsy - Flight over Rio (MP3)
===================================================
WXP  / FB2k :         -3.95, 1.029111, -3.40, 1.076560
LMDE / wine + FB2k :  -3.95, 1.029111, -3.40, 1.076560
LMDE / MP3Gain :      -4.75, 1.029080, -4.20, 1.076527
LMDE / SndC + FB2k :  -3.94, 1.000000, -3.40, 1.000000
LMDE / SndC + mtfl :  -4.75, 1.000000, -4.20, 1.000000

Al Di Meola - Elegant Gypsy - Elegant Gypsy Suite (MP3)
=======================================================
LMDE / wine + FB2k :  -4.42, 1.076560, -3.40, 1.076560
LMDE / SndC + FB2k :  -4.42, 1.000000, -3.40, 1.000000
LMDE / SndC + mtfl :  -4.68, 1.000000, -4.20, 1.000000


LateNightTales - Cinematic Orchestra - Ritournelle (FLAC)
=========================================================
WXP  / FB2k :         -7.86, 0.909882, -4.88, 1.000000
LMDE / wine + FB2k :  -7.86, 0.909882, -4.88, 1.000000
LMDE / mtfl :         -6.11, 0.909882, -5.13, 1.000000

LateNightTales - Cinematic Orchestra - Auntie's Harp (FLAC)
===========================================================
LMDE / wine + FB2k :  -3.48, 0.991547, -4.88, 1.000000
LMDE / mtfl :         -4.47, 0.991547, -5.13, 1.000000

Legends
=======
WXP :  Windows XP SP3
LMDE : Linux Mint Debian Edition
wine : Windows on Linux 1.0.1

FB2k : Foobar2000 1.1.6
MP3G : MP3Gain 1.5.1
mtfl : Command-line FLAC metada editor 1.2.1

SndC : Gnome SoundConverter 1.4.4
flac : Command-line FLAC encoder/decoder 1.2.1


Conclusions
  • Foobar2000 on Windows and on Linux gives the same results. So, wine doesn't alter the computation.
  • Reincoding the original MP3 file to FLAC doesn't changes the gains but do change the peak values. Why is that?
  • There is a difference between the gains computed by Foobar2000 and MP3Gain/MetaFLAC. In my test run, it can be as big as a 1.75dB difference! Why is that?


Can anybody explain those differences? What software is the more precise?

Regards,

Olivier

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #1 –
assuming foobar 1.1.6 or later:

http://www.foobar2000.org/changelog

Quote
ReplayGain scanner now uses libebur128 for improved accuracy.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #2 –
I guess you did not read the foobar2000 change log. The Replay Gain system has changed. It uses now a different algorithm. I don't know if the new system is better or worse than the old.

A quote from the change log:
Quote
1.1.6
ReplayGain scanner now uses libebur128 for improved accuracy.


EDIT

marc2003 posted  while I was copying the quote...

IMHO, "improved accuracy" is unproven. Most of the discussion here has been about the technical implementation.
It would probably be very difficult to organize a comprehensive listening test.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #3 –
Quote
Reincoding the original MP3 file to FLAC doesn't changes the gains but do change the peak values. Why is that?

The peaks that go over 0 dBFS (aka 1.0 in foobar2000) are truncated because the PCM content is not stored in a float format in FLAC files. Usually this kind of "lossy" peaks exist only here and there and have very short durations (only a few samples for each) and thus they do not have a significant effect to the measured replay gain value.

EDIT

By "lossy" peaks I mean the increased peak values that are byproducts of the lossy encoding/decoding process. They are one kind of lossy encoding artifacts, though usually inaudible even when brutally truncated because of their extremely short duration.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #4 –
Quote
IMHO, "improved accuracy" is unproven.


Makes me worry since I have been waiting for foobar2000 1.1.7 stable, since I will re-scan my entire library.
 

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #5 –
There are 100's of posts in the appropriate threads going over every nuance of this issue. What is there to worry about?
Leave your files as they were or re-scan - meh.

For the vast majority of tracks, if not your whole library, you not notice a difference. That is, you will not hear a difference. The numbers may change.

As before, there will always be Edge cases where the gain is noticeably not be where you would like it. Which files are edge cases may or may not change.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #6 –
Quote
What is there to worry about? (...) - meh

Unproven "improved accuracy" is a strong statement, that's all.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #7 –
Thanks for the infos.

I didn't read the release log of Foobar2000 for a long time. Mea culpa.

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #8 –
Quote from: bibmad on
I ran some tests with different softwares and found that the replaygain values may significantly differ. (...)

I've had the same problems.  I think someone mentioned in the thread that you can't hear the difference.  Maybe I'm extremely picky, but I can definitely hear the difference when wearing headphones.  I need a more precise software too.  Has anyone found a better software for reincoding?

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #9 –
Quote from: whitejason on
Maybe I'm extremely picky, but I can definitely hear the difference when wearing headphones.  I need a more precise software too.


Then rescan with the updated foobar.

Quote from: whitejason on
Has anyone found a better software for reincoding?


What are you reencoding exactly?  Could you explain what you're trying to do, would make it easier to answer your question

ReplayGain: Foobar2000 results differ from MP3Gain and MetaFLAC ones

Reply #10 –
Quote from: saratoga on
Quote from: whitejason on
Maybe I'm extremely picky, but I can definitely hear the difference when wearing headphones.  I need a more precise software too.


Then rescan with the updated foobar.

Quote from: whitejason on
Has anyone found a better software for reincoding?


What are you reencoding exactly?  Could you explain what you're trying to do, would make it easier to answer your question

Sartoga,

Thanks for your response.  I rescanned with the updated foobar.  It sounds a lot better.  Maybe it was just my headphones playing tricks on me.  I tried a new pair of headphones.  Between the new headphones and rescanning with the updated foobar, I can tell a big difference.  Disregard my second question.  I don't think I will need another software now that I have this all figured out.

Thanks again Sartoga!