Does enhancing MP3 files through editing actually improve it?

Hi,

I was wondering, if one had, for example, an MP3 file at 192Kbps and then used software such as Cool Edit Pro to modify it--get rid of clips, neutralize, add this and add that, in order to make up to some degree what the compression of the MP3 file did to the original, Would it in fact improve the original? I know 'improving' is based on the 'ear of the beholder' so to speak but would it sound better enough so that this same 192 Kbps file be then saved to 224Kbps (or 320) since it was modified and 'improved'?
Or is saving to a higher Kbps actually useless (or makes it worse)?

thank you

No.

When a file is saved in MP3 format, a lot of bits and bytes are removed from the file (MP3 is a lossy compression method, after all). On a decent set of speakers you can hear this. It is impossible to add information to the MP3 by saving it in a higher bit rate. What's gone is gone.

It only gets worse when you save it as a MP3 again: AA uses its own file format for editing, and when saving as an MP3, it compresses the file to the lossy format, thus causing (again) a decrease in quality.

The actual fact of the matter is that while there is a resolution reduction in encoding to mp3, if it is done well, it works. What "works" means is that mp3 is perceptual encoding with the goal of saving storage space and transmission time without audible degradation. This is not hard to do (using existing tools)  in spite of the fact that there are many poorly done mp3s to hold up as examples. Also, there are many low bitrate mp3s created where something else was much more important than ultimate audio quality.

Well done mp3 (e.g. Lame V2 VBR encoding (actually even V5)) is transparent for most music for most people. There are always exceptions. The statements about good quality apply to "most music for most people," but most people who believe they can tell the difference between uncompressed and good mp3 have not done ABX comparisons.

Re-encoding after editing goes through the same process without regard to the fact that the file started as an mp3. In theory there is probably measurable change, but audible changes are another matter. One re-encoding, to the same or a higher bitrate or quality setting, is not likely to be noticeable or ABXable.

Therefore, if the audio need attention, such as a raw LP transfer, it can definitely be improved by decoding to wave, cleaning it up, then re-encoding -- after all changes are completed.

Therefore, if the audio need attention, such as a raw LP transfer, it can definitely be improved by decoding to wave, cleaning it up, then re-encoding -- after all changes are completed.

True, but the question was specifically on enhancing and recoding of MP3 files.

I do agree with you that most people, who think that iPod and the likes are the standard for music quality, will not notice any difference between CD quality and MP3 on 192 kbit or higher. There's nothing wrong with that. What bothers me is that people take a lower quality for granted.

If the question relates to audio problems due to the initial mp3 encoding, then I think you are correct, it is unlikely that CoolEdit can improve it. This has nothing to do with whether one wishes to then re-encode to mp3 again, which is what I was addressing. It is rather that the kinds of changes mp3 makes can't really be reversed. It those changes are detrimental to audio quality, you are stuck with them, just as you are stuck with most general distortion from whatever cause.

The simple answer is that bringing an mp3 file into any wav editing program and then re-saving as an mp3 only throws away more of the data that was left over from the last mp3 encoding.  Making it a wav does nothing to improve or add to it.  Every mp3 save is discarding more of the original file.  You only get degradation.

The simple answer is that bringing an mp3 file into any wav editing program and then re-saving as an mp3 only throws away more of the data that was left over from the last mp3 encoding.  Making it a wav does nothing to improve or add to it.  Every mp3 save is discarding more of the original file.  You only get degradation.

But if the original file wasn't very good (before compression) you may be able to do some processing in between that improves the perceived quality more than loss caused by the second compression.

Paul

"Perceived quality" is the right phrase here. If people are happy with what they hear and what they think/perceive is good quality, then who are we to tell them differently. Sometimes we have to tell them it can be (way) better, and that's where the opportunities are for us, the Audition People.

Thanks ye all on your advice
--that is why, on a different post, I had asked if it were possible to edit MP3s losslessly. I thought that since a program like 'MP3 Gain' exists whereby gain is changed losslessly then, in theory at least, editing could be as well.
I am of the type who when dealing with the original will convert to MP3 320kbps (or other formats. It is because so many people (Aaargh) compress to 192kbps or less that drives me bananas--and do so badly--whereby because of one peak the file is too low (sounding). In my experience MP3 gain is great but it reads that one peak as well and therefore, often enough, nothing can be done to the file. That's where I take out CoolEdit in order to 'fix' the problem. In my case sometimes I do hear a good enough change, but often enough the file is worsened. Because editing takes time ( a lot of it at times) it pisses me off to have done all that work for nothing. It is a gamble every time I do so--will it improve or not? hence my question.
Thanks again

I’ve done a fair amount of editing from mp3 source, then re-encoding, without being able to find audible problems. It is too obvious a question, but are you certain that whatever problems you find are not there before making the mp3? A bitrate of 192, especially if it is the average for VBR, can do a pretty good job. I would look at the mp3 encoder and settings you use.

AndyH  Sorry for not being specific. I was not referring to mp3 files I make but rather those done by someone else. As mentioned mine are always 320kbps and I was referring to those of 192 or lower.

Keep in mind that VBR files will have "busy" frames encoded at 320.  I have a fairly good setup & ears, yet can't hear the difference between a 320 encode and a LAME VBR encode on "standard" - which averages around 175k, IIRC.  I think if you do an ABX you may be surprised.  Of course, if others make gross errors in editing or encoding, the bitrate becomes moot.

thanks to all

I never use VBR. I find the kbps always come out too low. I once did an experiment whereby I chose a popular song and downloaded it from everywhere. I then did a comparison I had three 320KBPS which sounded best of all (one 320 wasn't as good). One 192 kbps that sounded quite good and the rest (192, 128) that sounded terrible--The biggest difference in quality was between the 128s--which were most common and the VBR ones were mixed--some good some terrible.
Of course this doesn't say much about how an MP3 can actually sound like if done properly. I would assess that the reason why the 128 and 192 didn't often sound good is because those who did them didn't care or didn't know what they were doing or both.

BTW what's ABX again (I'm at an airport and don't have my software with me)

Thanks again

M

ABX is a blind A/B comparison. It goes on the same basis as all medical drug testing, the same basis as any and every scientific investigation of human response to just about anything.

Here is A (e.g. the uncompressed wav).
Here is B (e.g. the LAME VBR V2 version of same).
Here is X, either A or B, chosen at random.
Your job is to decide which X is. You do ten or more trials on a sample, then see the results to find out how often you were correct. On line are various detailed discussions of the reasons for doing this and of the statistical evaluations of the tests. http://www.hydrogenaudio.org/forums/index.php?showtopic=16295&

There is a variety of freeware that makes this task simple. By now many millions of people have been presented with the hard evidence that what they often believe are obvious audible differences are only self created delusions. Expectation and belief are very strong factors influencing our perceptions in many circumstances. When we have no information except what we actually hear, we may find that no big difference exists, that no audible difference at all exists, or we may even find that there is some undefined difference that none-the-less allows us to make the correct identification.

WinABX http://www.kikeg.arrakis.es/winabx/
PCABX http://www.pcabx.com/
and an ABX plugin for foobar2000
are three of the more popular ABX tools. WinABX and PCABX do not do real time decoding. If mp3, or some other compression codec, is part of your test, you must first decode to wav and do the test on those files. This is identical to playing the compressed version in any software or hardware player.

[know]

The simple answer is that bringing an mp3 file into any wav editing program and then re-saving as an mp3 only throws away more of the data that was left over from the last mp3 encoding.  Making it a wav does nothing to improve or add to it.  Every mp3 save is discarding more of the original file.  You only get degradation.

While I do not doubt that (a) you can't put back what has already been discarded and (b) there is most likely going to be a generational loss going from MP3 generation 1 to MP3 generation 2 (plus some further), I do not know for sure that "Every mp3 save is discarding more of the original file".  I have, in fact, wondered whether or not this is the case.  I have not done the tests myself, nor am I familiar enough with the psychoacoustic algorithm that decides what frequencies to throw away to know whether or not there is a limit beyond which the algorithm (at whatever chosen bitrate) realises there is no more information to throw away.

The reason I question this, is because the JPEG algorithm works in a similar way: it discards picture details to achieve compression at some chosen compression level.  And I have done this test with JPEG: repeatedly save generations of an image as JPEG, in the expectation that it would eventually become a featureless blur.  But the image didn't become a featureless blur: it settled onto a very blocky representation of the original scene, but further generations were bitwise identical.  The algorithm had clearly thrown out all it was ever going to throw out, and settled on the minimal data at the chosen quality level.  I wouldn't be surprised if the same thing happened with MP3.

Not that any of this helps the original poster.  You will lose more than you gain, and you only stop losing when the generation is so high you have already lost much detail.  But, pedantically speaking, I'd be surprised if you only ever get degradation with every generational save (otherwise you'd end up with effectively zero-length encoded data, with no content whatsoever).