Removing clipping from an MP3 file

Audacity 3 Help Forum Windows
1

Howdy

New to this forum, but anyhow, last year, I recorded a file which has pretty bad clipping in it. It’s not that bad in which one can’t hear any of it clearly at all, but it’s certainly not clear of clipping. I was just wondering, are there any tools in Audacity that can eliminate this clipping? I’ve tried the de-clip plugin, and I can’t seem to get it to work properly. Basically it I just waiting for the process to end, and it’s exactly the same afterwards. Basically, I just want to replace the faulty bits with stable ones, but I’d like to do it using a in a program, instead of a long daunting task doing it manually.

I’m using version 2.1.0 btw. Windows 7 Home Premium 64 bit

Here’s a preview of the bad file. I really want to get this one fixed if possible, as the actual audio its self stands out to me like dogballs. It’s been annoying me for months now, but I recorded it with the wrong settings unknowingly.

http://i7.photobucket.com/albums/y255/SAR704/0805cprev_zps2ncs8suu.jpg

Thanks

2

That waveform doesn’t look clipped. At least not badly… There are couple of negative spikes, but at that zoom level it’s hard to know if there’s some kind of glitch, or clipping, etc. And in general, you can’t judge sound quality by looking at the waveform…

A couple of times when I tried Clip Fix, it made the waveform look better, but it didn’t fix the sound.

If you select View → Show Clipping you should see red where there’s clipping, and if you zoom-in you should see [u]clipped[/u] waveforms. (You might have to use the Amplify effect to reduce the volume in order to see the clipped peaks.)

Note that MP3 compression often boosts some peaks by one or two dB . MP3 files can to over 0dB, and this peak-boost can result in Audacity showing potential clipping even though the wave peaks are not clipped.

3

When rendering to an mp3, the peaks should be below -3 dB. If not, mp3 will sound awful. Even without clipping…

If you really need to compress dynamics, use a multi band compressor if your end result will be an mp3. It could also be wise for other lossy compression, but I’ve never experimented a lot with those.

There’s no real fix after the facts. That’s why you should at least keep an uncompressed archive copy of your project, so you can go back.

There are some expensive plugins that can make clipping less noticeable, but I have no idea if any of them work with Audacity, as I don’t own them…

4

Does EZ-Patch help?

Gale

5

When rendering to an mp3, the peaks should be below -3 dB. If not, mp3 will sound awful. Even without clipping…

I’ve never heard of a case where reducing the level before encoding prevented or reduced compression artifacts.

NOTE - A proper scientific, blind, [u]ABX Text[/u] must be level matched. That means you’d have to reduce the loud MP3 (after encoding) before comparing the sound quality of the loud & quiet versions. And, you’d have to do that with a tool like MP3directCut that allows level changes without de-compressing/re-compressing.

6

I’ve put quite some time in researching why some of my mp3’s in iTunes sounded so bad.

First, I suspected bitrate. But then i found some lower bitrate mp"s sounded good.
Then I suspected the software used to convert to mp3. There was no pattern to be found.

So I started experimenting with CD’s and ripping them with iTunes and other encoders. No pattern either. Until I ripped a Metallica CD that was compressed to zero dB and already had some clipping. It was awful, once ripped.

So I took some very good quality productions from Zappa, Chris Rea and others and started amplifying those before ripping. No problem until you go over -3 dB.

Long after that, I found a youtube vid from TC Electronic in Denmark. It was about the new LUFS metering system for broadcast and how that would kill the incentive to over compress. And there was a specialist from TC Electronic confirming what I discovered on my own.

Lemme see if I can dig up the vid. It’s very enlightening on the subject of compression.

NOTE - A proper scientific, blind, > [u]ABX Text[/u] > must be level matched. That means you’d have to reduce the loud MP3 (after encoding) before comparing the sound quality of the loud & quiet versions. And, you’d have to do that with a tool like MP3directCut that allows level changes without de-compressing/re-compressing.

I know about levels and have the software to do blind listening tests. I wouldn’t use a tool like MP3directcut, as you can easily adjust playback level in the DAW when necessary.

When I started all of this, I started with the suspicion that my old monitors needed a rebuild. In fact, I rebuilt some of them, to be quite disappointed. The mp3’s even sounded a bit worse, afterwards…

I also don’t ever do blind listening tests on my own. We do these with at least five listeners, all of them audio engineers or at least people who have listening experience and can compare with live (classical) music. One person rules the button that switches, the four others listen and take notes. Of course, we rotate the person who controls the button… :smiley:

7

Here’s the video:

https://www.youtube.com/watch?v=cs0Vq9XrT2U