Double compression, mp3 to AIFF and back to mp3 Options

[shenzi]

I have a number of mp3 files (and Ogg Vorbis) with dropouts in them from recording via the web (Radio7 comedy, that sort of thing). If I import the file into Audacity and repair the break and then export again as mp3 at the same bitrate as the original ... is this the best way to do this?

I conscious that saving and resaving JPEG images throws away data, so am I doing the same with the mp3s? (I haven't tried a listening comparison yet but I'm interested on a technical level anyway.)

[Nick.C]

If bitrate is not a problem, why not convert to WAV / AIFF then save to FLAC? In this way you will lose nothing more.

[rpp3po]

In my opinion saving mp3 quality at FLAC bitrates is a huge waste of space. Reencoding back to mp3 at the same bitrate is a huge waste of quality.

Better try one of the lossless mp3 frame cutting tools and cut out the artifacts.

[Axon]

Have you tried mp3DirectCut?

[shenzi]

Have you tried mp3DirectCut?

No but I bet it's a PC program, right? (Excuse me whilst I Google ...). The reason for saving back as mp3 is to put on a CD for use in the car. Also it seems wrong to expand a file to fill hard disk space ...

Rats, yes, PC and Linux. Looks a neat program though.

[DVDdoug]

I conscious that saving and resaving JPEG images throws away data, so am I doing the same with the mp3s?

Yes. Conceptually, it's the same. I've never done any experiments or ABX tests either. (Nor have I done these experiments with JPEGs.)

Has anyone done any ABX tests to see if a 2nd encode (at the same bitrate) actually makes a difference? If somebody doesn't already have the answer, maybe I'll put that on my "rainy day" list. I think it would be most interesting to do this just-above and just-below the transparancy threshold.

I try to avoid re-coding/transcoding lossy formats, and I generally consider it a "bad" thing... But it might not be so bad... MP3 tries to throw-away the least-audible information. During the 2nd encode, most of the "information" it would throw-away isn't there anyway... Does that make sense? During the second encode, it's starting-out with a less-complex AIFF file, and less "real" information needs to be thrown-away to achieve the target bitrate.

I would assume that transcoding from OGG to MP3 might be more-likely to introduce artifacts, since they are likely to throw-out different information. If I ever have to do this, I'd probably use a higher bitrate for the 2nd format, just to increase the safety margin.

[rpp3po]

2 cents:

1. Transcoding artifacts for popular audio codecs are not anything only perfectionist would notice. Just give it a try, it's really bad, even for high bitrates.

2. Strangely transcoding from LAME to LAME produces much worse results than transcoding between different formats. I don't remember the thread or wether they have found an explanation in the end. Just use the search function, if you want details.

[Ardax]

No but I bet it's a PC program, right?

Rats, yes, PC and Linux. Looks a neat program though.

Just a little bit more googling would have revealed that at least some people are successfully running it under Wine, and there's mpcut for Linux.

[shenzi]

Okay, Googled Wine and (after a few splendid offers from my local merchants) found out more about it. I don't thin it's the sort of thing I'd get involved with to translate a few mp3 files. (I'm Mac. remember, no great desire to wander out into the big, wide world.)

Thanks for the comments so far. So theoretically, would the best route be to go mp3 -> AIFF (or smilar non-lossy format) -> mp3? I assume Audacity's native file format is non-lossy. This is similar to sort of route I would take if editing a JPEG via Photoshop.

[rpp3po]

Thanks for the comments so far. So theoretically, would the best route be to go mp3 -> AIFF (or smilar non-lossy format) -> mp3?

So practically, after all this advice, you think that the theoretically worst route is the best.

[kornchild2002]

Also don't forget that your source (such as an internet radio) was also lossy. So you have already gone the lossy to lossy route. You would basically be re-encoding things for a third time. Many online radios use AAC, HE-AAC, WMA, and/or mp3 when streaming their content. You recorded that and made your own mp3 files. Down you want to edit those mp3 files and then re-encode them for a third time using a lossy encoder.

I guess you could try it and see if you could hear (that doesn't sound right) the artifacts introduced from lossy encoding the audio. It might be hard to spot the artifacts if the recordings are voices, I think it would be relatively easy if the recordings consist of music.

[shenzi]

So practically, after all this advice, you think that the theoretically worst route is the best.

Ah no, I thought I had two sets of conflicting advice. Perhaps my mistake. It was easier when it was just razors and quarter inch tape.

However unless I use an mp3 editor of some sort like the one mentioned above, I am going to have to come out of one format and save back into it in order to do the edit. Since starting this thread, I've tried it with one music track using Audacity. The end result sounds similar, albeit slightly louder. Whether that means it's reduced the dynamic range, I can't say, but it doesn't seem to have introduced any unpleasant artefacts. But I need to have a good listen at home rather than in the car. I'll also play with Garageband and see if that has any facility to do this.

[Ardax]

I'm Mac.

That might have been handy to know in the opening post. When you said "PC and Linux" above, I thought that's what you were running.

Have you found MP3 Trimmer yet? It's my first Google result for "mp3 edit Mac", and looks like exactly what you're looking for, at least for the MP3s.

[shenzi]

Excellent, thank you.