Help - Search - Members - Calendar
Full Version: How to prevent transcodes + 2 questions
Hydrogenaudio Forums > Lossy Audio Compression > MP3 > MP3 - General
ec461
Hi,

I had a suggestion for the LAME codec. Many times users transcode files from a lower bitrate to a higher bitrate thinking they'll get better quality. Of course, these people are idiots but... Why can't LAME simply detect the exitsing bitrate and then refuse to transcode to a higher bitrate? That way ppl simply can't create a transcode. Of course, you can still create a wav and then transcode it..but I'm not sure how many ppl will do that. (or maybe LAME could even put a header in the WAV file saying "created at x bitrate")

Q1. Is aps/V2 *excatly* the same as 192VBR? and apx/V0 *exactly* same as 256VBR? If not, why?
Q2. If I transcode from a higher bitrate to a lower bitrate (lets say APX to APS), will the resultant file be exactly the same as the file which is directly encoded from the original source at APS? Again, if not, why? I mean after all APX has more info than APS and so all LAME has to do is remove this extra info, right?
gfngfgf
Q1: I'm not sure what you mean by "192VBR" and "256VBR". Maybe I'm just tired right now.

Q2: No, not at all. The problem with your logic is that you have two different starting points, so to speak: one is the original file, the other is an APX-encoded file. Because these two files are different, the analysis LAME performs will give different results, and as such, the compression will end up sounding different.
ec461
QUOTE(gfngfgf @ Nov 26 2005, 12:51 PM)
Q1: I'm not sure what you mean by "192VBR" and "256VBR".  Maybe I'm just tired right now.

Q2: No, not at all.  The problem with your logic is that you have two different starting points, so to speak: one is the original file, the other is an APX-encoded file.  Because these two files are different, the analysis LAME performs will give different results, and as such, the compression will end up sounding different.
*



I mean 192kbps VBR and 256kbps VBR.

Well what would it take for LAME to encode information so that it can be 'stripped away' when transcoding from a higher bitrate to a lower one? What's the difficulty in implementing this feature? Does it require a change in the whole architecture/algorithm of Mp3 encoding?

If that doesn't work can't mp3 have a correction file like WavPack? What this file does is that it converts a lossy file back into a lossless file. So, say you have a 192 kbps CBR mp3 file, can't you have a correction file for 224 kbps, 256kbps, and so on?

I'm just thinking aloud. I have no idea how the mp3 algorithm works apart from that it throws away certain frequencies to create perceptual transparency.
Mike Giacomelli
QUOTE(ec461 @ Nov 26 2005, 12:31 AM)


I mean 192kbps VBR and 256kbps VBR.

*



Ok you've narrowed it down to about 4000 possible combinations. Can you be more specific? 192 VBR encoded how? Depending on what you use, it may or may not be APS.

QUOTE
Well what would it take for LAME to encode information so that it can be 'stripped away' when transcoding from a higher bitrate to a lower one? What's the difficulty in implementing this feature? Does it require a change in the whole architecture/algorithm of Mp3 encoding?


What you're refering to is called bitrate peeling. Its been attempted in Vorbis which is supposed to support the feature. Unfortunately, its evidently nearly impossible to get it to work well. AFAIK no one has ever written a peeler that was even close to the quality you get just from transcoding.

QUOTE

If that doesn't work can't mp3 have a correction file like WavPack? What this file does is that it converts a lossy file back into a lossless file. So, say you have a 192 kbps CBR mp3 file, can't you have a correction file for 224 kbps, 256kbps, and so on?

I'm just thinking aloud. I have no idea how the mp3 algorithm works apart from that it throws away certain frequencies to create perceptual transparency.


You could write something like that. The correction file would probably be uselessly large, but it could be done. I wouldn't bother though. FLAC would be smaller and easier.
Never_Again
In addition to what Mike said:

there is no 192VBR (or 256VBR or any other <insert figure>VBR), for the simple reason that VBR targets quality not bitrate. Depending on the audio content, files encoded at -V2 can have average bitrate of as low 128kbps (e.g. mono material) and as high as 270kbps (e.g. metal).

As for the bitrate peeler, there is a MP3 implementation by Omion that can reduce the size of --preset insane files by up to 10% losslessly.
ec461
Ok I understood everything now.

I was wrong about the bitrate setting for VBR files, I didn't realise that it was based on quality.

Thanks for your replies.

What do you think about my suggestion for preventing transcodes from lower-->higher bitrate?
ShowsOn
QUOTE(ec461 @ Nov 27 2005, 02:26 PM)
What do you think about my suggestion for preventing transcodes from lower-->higher bitrate?
*



I think what you are refering to is called "bit rate peeling" this is not a part of LAME, but it is possible in other formats. For example Ogg Vorbis has this capability, but I do not think anyone has written an encoder that can actually perform the peeling.

For me it wouldn't be worth it. I struggle ABXing MP3 files made using LAME 3.97b1 using the "-V4 --vbr-new" setting.

For my iPod I use iTunes 160K VBR AAC, and am yet to succesfully ABX a file encoded with that setting. Crap hearing has enabled me to keep some free space on my iPod :-)
Lyx
You cannot prevent it, because in most cases, LAME will not see an MP3 as the source, but a WAVE....... when you reencode an MP3, then most tools will do the following: MP3 ---decode--> WAVE --encode(LAME)--> MP3

- Lyx
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.