Useful tip to fix distorted (overly loud / poorly remastered) albums

[WinterRacer]

Sorry if this has been posted before or is in the wrong section, but I thought it a tip talked about on the AVI forum recently is worth sharing.

I've bought some albums recently, that are casualties of the loudness war with obvious disortion - so much so I found I really didn't enjoy listening to them. Lisa Hannigan's 2nd album, Passenger, is one that springs to mind, especially disappointing as her first was so well produced.

Basically, a programme called Audacity has the ability to show clipped peaks and fix them.

Here's an image of a track, with clipped peaks shown in red:

The same track after I've fixed it and sounds much better.

The workflow to do this is

1. Download and install Audacity (it's open source so free)

2. Convert you track into a format Audacity can work with, e.g., mp3 or FLAC

3. Reduce volume by 10db - Select "Effect" then "Amplify..." from the menus and enter "-10db".

4. Select "Effect" then "Fix Clip..." from the menu

5. Select "Effect" then "Amplify..." from the menu and leave the volume to amplify as the volume Audacity choses for you.

6. Select "File" then "Export..." to save the track in the format of your choice.

Of course, we shouldn't have to do this, but if you've got some albums that suffer from the loudness war, this might make them listenable again.

 

[kevinJ]

I'm going to test this myself.

But I don't know what to think of it. If you change the dynamic range you just add some more compression/limiting stuff. Or something like that.

 

[ESP2009]

I will be interested to see what people have to say on this one. I had not knowingly come across serious casualties of this so-called 'war' previously. However, the other evening I decided to play a couple of tracks I had ripped to flac format from the Beatles compilation album simply entitled '1'. It had languished in my CD cabinet for a few years, unheard. Now I know why. Yikes! Talk about horribly bright and shouty! Yeugh! The classic 'Something' from Abbey Road was unlistenable. For comparison I loaded the 'original' ripped from the actual Abbey Road album and there was a world of difference. Infinitely preferable and easy on the ear.

If Audacity can 'heal' such woes I am all in favour of giving it a try.

 

[Paul.]

I'd like to ask a very slightly off topic question just to clear something up in my own head. The term clipped is used in photography and video when the exposure exceeds the dynamic range abilities of the camera, and the histogram chart displays a flat peak at the top of the graph. I assume the same thing is happening here, if the audio is clipped then the volume has exceeded the dynamic range of the recording?

If this is the case, how can this be repaired? In photography, when clipping occurs, no detail data is recorded, usually just solid 255.255.255 white (depending on colour temp of light source) and nowt else. Surely all audacity is doing is reducing the volume, but the flat top is still present from lack of data? Surely this data is still clipped? I can reduce the sample volume in itunes by album much easier than this process, so whats the difference?

 

[ESP2009]

Fuzzy logic extrapolation algorithm? Sounds good to me.

 

[The_Lhc]

If this is the case, how can this be repaired? Surely all audacity is doing is reducing the volume, but the flat top is still present from lack of data? Surely this data is still clipped? I can reduce the sample volume in itunes by album much easier than this process, so whats the difference?

I had exactly the same thought myself. Be afraid Hobbs...

 

[WinterRacer]

I'm going to test this myself.

But I don't know what to think of it. If you change the dynamic range you just add some more compression/limiting stuff. Or something like that.

When you look at the waveform you should see nice rounded peaks (think sine wave), but when digital clipping occurs, these rounded peaks are squared off, e.g., 5 consecutive samples at 0db (full volume in the digital world).

What you do with Audacity is first reduce the overall volume. This is to allow somewhere for the new reconstructed peaks to go. Next, Audacity analyses the track and looks for squared off peaks. Using the samples before and after the squared off peak it guesses what the peak should look like.

After this is done, you amplify it again, to get the new peaks back up to 0db level.

It's not adding more compression, quite the opposite. Does that make sense?

 

[WinterRacer]

I will be interested to see what people have to say on this one. I had not knowingly come across serious casualties of this so-called 'war' previously. However, the other evening I decided to play a couple of tracks I had ripped to flac format from the Beatles compilation album simply entitled '1'. It had languished in my CD cabinet for a few years, unheard. Now I know why. Yikes! Talk about horribly bright and shouty! Yeugh! The classic 'Something' from Abbey Road was unlistenable. For comparison I loaded the 'original' ripped from the actual Abbey Road album and there was a world of difference. Infinitely preferable and easy on the ear.

If Audacity can 'heal' such woes I am all in favour of giving it a try.

If you've got a PC it's free. First thing to do is load the track, select "View" and then "Show Clipping". This will indicate how bad the clipping is.

 

[WinterRacer]

I'd like to ask a very slightly off topic question just to clear something up in my own head. The term clipped is used in photography and video when the exposure exceeds the dynamic range abilities of the camera, and the histogram chart displays a flat peak at the top of the graph. I assume the same thing is happening here, if the audio is clipped then the volume has exceeded the dynamic range of the recording?

If this is the case, how can this be repaired? In photography, when clipping occurs, no detail data is recorded, usually just solid 255.255.255 white (depending on colour temp of light source) and nowt else. Surely all audacity is doing is reducing the volume, but the flat top is still present from lack of data? Surely this data is still clipped? I can reduce the sample volume in itunes by album much easier than this process, so whats the difference?

The first part you're right on. However, the clever bit is the "Clip Fix", Audacity uses the samples before and after the squared off peak to interpolate what the peak should look like. It's not perfect, but it's made some albums that sounded horrible, sound a lot better.

 

[fr0g]

It definitely works, but is a pain in the bum to do. They really need a batch version.

Of course it can't completely fix some of the utter rubbish out there, but it has an effect of reducing the obvious distortion.

And there's nothing it can do to fix Justin Beiber records I'm afraid.

 

[WinterRacer]

I'd like to ask a very slightly off topic question just to clear something up in my own head. The term clipped is used in photography and video when the exposure exceeds the dynamic range abilities of the camera, and the histogram chart displays a flat peak at the top of the graph. I assume the same thing is happening here, if the audio is clipped then the volume has exceeded the dynamic range of the recording?

If this is the case, how can this be repaired? In photography, when clipping occurs, no detail data is recorded, usually just solid 255.255.255 white (depending on colour temp of light source) and nowt else. Surely all audacity is doing is reducing the volume, but the flat top is still present from lack of data? Surely this data is still clipped? I can reduce the sample volume in itunes by album much easier than this process, so whats the difference?

Hopefully what Audacity is doing is recreating the Red parts of the signal from the surrounding data and thus may recreate peaks above the flat-top. I'm sure it could do this if such a feature were coded (I'll have a try with some C# soon to see what mess I can make), but I'm not sure if the steps outlined by the OP actually does that. I think is it just decompressing the top 10 db of the signal , will have to play with it this w/e also.

I'm sure there are programs out there that also fix the flat-top. Could be a fun project though.

EDIT: Just saw the reply above come in before mine - so it does do clip fix - nice! And does it decompress also? Yes it does "Select "Effect" then "Amplify..." from the menus and enter "-10db" -- I should read better!

-N-

The algorithm is available from the 'plug ins' directory of Audacity, but you'll have to convert from LISP to C#. What would be useful is a macro (chain in audacity speak) to do this as an action on a bunch of files. File by file is a bit tedious.

 

[Anonymous]

I'd like to ask a very slightly off topic question just to clear something up in my own head. The term clipped is used in photography and video when the exposure exceeds the dynamic range abilities of the camera, and the histogram chart displays a flat peak at the top of the graph. I assume the same thing is happening here, if the audio is clipped then the volume has exceeded the dynamic range of the recording?

If this is the case, how can this be repaired? In photography, when clipping occurs, no detail data is recorded, usually just solid 255.255.255 white (depending on colour temp of light source) and nowt else. Surely all audacity is doing is reducing the volume, but the flat top is still present from lack of data? Surely this data is still clipped? I can reduce the sample volume in itunes by album much easier than this process, so whats the difference?

Hopefully what Audacity is doing is recreating the Red parts of the signal from the surrounding data and thus may recreate peaks above the flat-top. I'm sure it could do this if such a feature were coded (I'll have a try with some C# soon to see what mess I can make), but I'm not sure if the steps outlined by the OP actually does that. I think is it just decompressing the top 10 db of the signal , will have to play with it this w/e also.

I'm sure there are programs out there that also fix the flat-top. Could be a fun project though.

EDIT: Just saw the reply above come in before mine - so it does do clip fix - nice! And does it decompress also? Yes it does "Select "Effect" then "Amplify..." from the menus and enter "-10db" -- I should read better!

EDIT again: Ok that only reduces the volume - I think instead of them fixing and amplifying afterwards, you may also want to put a touch of decompress in there too, this will be on a track by track basis unfortunately.

-N-

 

[Anonymous]

And there's nothing it can do to fix Justin Beiber records I'm afraid.

Really? Try this: http://gawker.com/5614579/how-to-make-justin-bieber-sound-incredible-slow-him-down-800-percent

-N-

 

[fr0g]

And there's nothing it can do to fix Justin Beiber records I'm afraid.

Really? Try this: http://gawker.com/5614579/how-to-make-justin-bieber-sound-incredible-slow-him-down-800-percent

-N-

Aha,indeed, I tried that last year. Very cool indeed.