Hello all,

I have a whole lotta MP3s which I want to transcode into Vorbis. Yes, I know it will further degrade the quality... but the amount of space I predict I will save is quite significant...

Well anyways , my plan is to make a table such as this one:


Plus another table comparing the file sizes.

The only limitation is that I have severely limited CD collection, so only some samples I can test with multiple MP3 bitrates. Others I have only the MP3s, and for these I test only the MP3's bitrate.

The process is like this:

(original wav) --> [LAME 3.97b2] --> (mp3) --> [LAME 3.97b2] --> (decoded mp3 wav) --> [oggdropXPd Lancer 20051121] --> (vorbis) --> [oggdropXPd Lancer 20051121] --> (decoded ogg wav)

Then I ABX-es the (decoded mp3 wav) and (decoded ogg wav).

Any input?

Yes.

Why bother having the decoded WAVs to compare between MP3 and Vorbis?
Just transcode the MP3 to Vorbis, and ABX them.
Decoding them to WAV to be compared seems like a waste of time and h/d space for me...

And if you plan to rip straight from an Audio-CD, try q 2.
It's quite space saving, and sounds great on noisy environs, even w/ earphones.

Foobar automatically does this (preparing files for ABX) for you. He's just replacing foobar (maybe he uses another script)

It's necessary to do this in order to not have timing / synch problems between the different format files.

As for suggested quality, It depends on your original bitrate and how much smaller you want to go.. but try -q 4, or -q 2 if you really need space.

I heard a -q 1.5 file today which I couldnt ABX to its original, as could not many other people on this forum; Try that setting if you like.

Peace,
Tristan.

Hmm...didn't know about that.
Though I still prefer comparing between the 'original' files in MP3 and Vorbis format.
Because, after all, those are what I'm going to hear to anyway, not the decoded WAVs.
But hey, if that statement's flammable then flame it.


Yup, I agree w/ you.
The last 80kbps test w/ previous aoTuV by guru at least'd give some ideas about the quality at that nominal bitrate.

-q 1.5 ... ha I didn't think of that. So, the q settings I will actually test goes like: -2, -1, 0, 1, 1.5, 2, 3, 4, 5, 6. I don't think above 6 is necessary... or should I also test them?

The reason I decode the MP3 to WAV is for my script to generate automatically all necessary .ogg files in one go. I save a lot of time decoding MP3 only once. The reason for decoding the .ogg files is to make it faster to compare between the q's... I think...

And yes I have been using q 2 for straight-from-CD-ripping... sounds real great even on the desktop (some artifacts sometimes show up, they are minor, and I can live with them )

Seems like u're easy enough to please. =)
In fact, -q 4 for transcoding from MP3 is probably enough for you.

But, you might wanna try this (not so) simple test.
Rip an Audio CD track into two MP3 files w/ different -V setting.
Say, -V 2 and -V 5 for around 256kbps and 128kbps, respectively.
Then transcode those two to Vorbis w/ -q 4 for each file, and ABX them.

If it's hard to ABX them, then -q 4 is probably your 'safe margin' for all of your transcode.

Just a thought.

In my superficial ABX tests on some killer samples some years ago, I found I had to use -q 6.0 before I failed to ABX against the original MP3. Sorry for the little details. I am not interested in transcoding any more.