I am looking for a software sound processor (either a tool or a
plugin, like a vst), that I would call an 'anti distortion' or 'anti
clipping' filter. I have some sound data that is subtly but audibly
clipping (distorting because it's recorded too loud).

Is there any (preferably cheap or free!) software out there that can reconstruct
clipped audio data? This may sound like a silly question (a big ask?),
but in theory I believe it to be possible, so I thought I'd ask...

By cheap I mean US$50-100 or so.

Try the demoware GOLDWAVE. it is a demo but i have used the demo
forever and it lets you have like 150 or so commands before you must
save your work and restart the program. thats unless you want to buy
it for 45$.

http://www.goldwave.com

best regards,
Solentech

Hmm - as far as I can see goldwave doesn't perform the function I'm
after (unless there is a plugin for it of course!). ;(

Thanks anyway,
Aidan

Sound Forge is professional editing software, and seems
to be capable of detecting and editing clipping:

Clipped Peak Detection and Marking
"Sound Forge software automatically scans sections of
 your file and marks sample peaks that are clipping.
 These indicators also show up in the Regions List for
 immediate identification. Audio Restoration tools
 eliminate clicks, pops, and other noise from your mix."
http://www.sonymediasoftware.com/products/showproduct.asp?PID=961&FeatureID=8106

The program costs ~$300, but also has a free trial download.

sublime1-ga

Thanks sublime -

I'm downloading the trial version now, but I'm sceptical that Sound
Forge performs the function I'm after ... :(

The sentence you quoted from their website seems close to answering my
requirement but it doesn't sound like they have an "Audio Restoration
tool" that actually reconstructs *clipped* sound data.

I'll post again if I find that you are right! ;)

Anyway, thanks again for your comment!

Hmmm -
I have downloaded and checked Sound Forge 8, but I can't find a way to
reconstruct clipped audio data with it. The closest I can get to
achieving this is using the "Audio Restoration" tool to remove noise,
but this seems to destroy the tone of the sample. I am not an expert
with the program, but it doesn't seem to have the function I am after.

Just to reiterate, I've drawn an ascii (yay!) diagram...
The hashes indicate data that has been lost during recording, that I
would like to have reconstructed:
                 ___
    ___         /###\
   /###\       /#####\
--/-----\-----/-------\------------
 /       \   /         \
/         \_/           \
                         \        /
                          \      /
---------------------------\----/--
                            \##/
                             ŻŻ

The way I expected a program to achieve this is to calculate a simple
cubic curve between the start and end of each clipped region. I would
have thought that using this technique, the difference between the
reconstructed wave and the original (unclipped) wave would have been
inaudible, at least for subtly damaged or simple waves.

Can anyone correct me, or should I just write a program/plugin to do this?

aidaan...

Sorry - I temporarily lost track of your question in the course
of a computer crash.

I think the Audio Restoration tool in Sound Forge 8 may be as
close as it gets to what you're trying to do, short of some
sort of audio reconstruction software used, say, by the CIA
and the like, which would cost a small fortune (or maybe a 
large one).

Given the nature of clipping, the data is essentially lost,
and while you might reconstruct it with averaging of the
remaining data, it will never sound quite the same. If
you succeed in creating a plugin that accomplishes this
to your satisfaction, it would likely be marketable.

Let me know...

sublime1-ga

sublime1-ga

Heh, yeah sublime I think you might be right - clipping *is* lost
data, and is probably unrecoverable. Perhaps my request was a little
naive!

I'll have a go at creating a simple reconstructor (using cubics),
which will probably close the book on that one (or open a can of
worms!).

Thanks for your comments - I'll post back here if after (if) I get
around to creating that program.

Because the sound information technically doesnt exist above
full-code,  you could try reducing the volume of the entire sound file
in Sound Forge,  then select the peaks that are gone one at a time and
use a command in SoundForge called "Replace".

This command will search the file for a fraction of a waveform that it
feels 'fits' that gap.

More often than not,  it finds seamless replacement data.  I use it
all the time when restoring badly recorded dialogue - that or draw in
the waveform manually.
Regards,

Neil Parfitt