Skip to main content
Topic: Transcoding Transparent -> Transparent (Read 2910 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Transcoding Transparent -> Transparent

While I understand in theory why transcoding lossy -> lossy is a bad idea, I'm very curious if there are any empirical results on the following scenario:  Assume you have a lossy file encoded using codec A, but at a bitrate that produces perceptual transparency compared to the original lossless file.  Then you transcode the lossy file produced by codec A into codec B, but at a bitrate that usually produces perceptual transparency relative to the source material.  In this case, this means codec B should produce perceptual transparency relative to its source, which is the output of codec A. 

Has anyone ever actually done any listening tests to prove that this kind of transcoding, where the results of codec A are transparent compared to the lossless source and the results of codec B are transparent compared to the results of codec A, can result in output from codec B that is not transparent relative to the lossless source in real world conditions, i.e. a reasonable number of generations and going from, say, MP3 to Vorbis?

Transcoding Transparent -> Transparent

Reply #1
I did some of these tests some years back, but nothing recently.
the losses do add up.
it may be impossible to tell the difference between generation n and generation n+1, but after a few iterations the differences from generation 0 (i.e., the original lossless file) will begin to be obvious.
God kills a kitten every time you encode with CBR 320

Transcoding Transparent -> Transparent

Reply #2
I've done it a few times and I had a hard time distinguishing any differences between the two.  This was usually with an mp3 to mp3 conversion. 

I only did this because I wanted one track from a cue sheet that used one mp3 for all tracks.

Transcoding Transparent -> Transparent

Reply #3
@JunkieXL: There are tools to cut (and even join) mp3 files

mpTrim, mp3directcut, to name a few...

@dsimcha: Yes, it has been tried a few times in these boards (don't have links at hand). Like it was said, the differences add up, and generally, it is worse to use the same encoder twice than using two different formats.


Transcoding Transparent -> Transparent

Reply #4
Has anyone ever actually done any listening tests to prove that this kind of transcoding

Some people like /mnt can ABX VBR0 (from the source) so you can imagine if it comes from transcoded files!
Personally, I'm sure I couldn't distinguish it. One thing would annoy me though: The fact that I know that I'm listening to transcoded files! 

SimplePortal 1.0.0 RC1 © 2008-2020