Poor Quality Interview Tape Recording

Hello everyone. I’m trying to use audacity to clean up a tape recording of an interview. For most of it I’ve managed to figure out what needed to be done, however I’m not sure how to eliminate (if possible) one problem that is pretty prevalent in the recording. I guess the mic that was used was complete crap because pretty much whenever this one person talks, it overloads the mic and you just get that horrible crackle/pop for the duration of him talking until his voice quiets some (I guess he has a loud voice to start with). Is it possible to eliminate this, or is it something that will just have to be dealt with? The recording is going to be used for dictation of the conversation, so I’m trying to clean it as best as I can w/o incurring further expenses.

Sorry, overload distortion isn’t fixable … https://forum.audacityteam.org/t/sound-engineering-help/12890/2

Version 1.3.11 has a “Clip Fix” effect that may mitigate the damage, enough that the person doing the transcription won’t get a headache.

– Bill

I had my expectations that overloading the mic wouldn’t be fixable. I will look into the clip fix, thank you both for your input.

magic

pls describe what this fix thing is doing
thanks

The best I can do is quote from the effect itself:

Clip Fix attempts to reconstruct clipped regions by interpolating the lost signal. Before use, reduce amplification by 10 dB to give room for the reconstruction (In other words, select the entire track and do Effect > Amplify with “Amplification” set to -10). ‘Threshold’ is how close to the maximum sample magnitude any sample must be to be considered clipped. If processing is slow, select only a few seconds of clipped audio at a time."

Clip Fix is a simple, stupid (but not blind) digital-clipping-corrector
The algorithm is fairly simple:

  1. Find all clipped regions
  2. Get the slope immediately on either side of the region
  3. Do a cubic spline interpolation.
  4. Go to next region

The default Threshold of 95% seems to work well.

To see it in action:

  • in a new project, Generate > Tone with default settings
  • select entire track
  • Effect > Wah Wah with default settings - clipping will be created
  • Effect > Amplify with Amplification -10 dB
  • Effect > Clip Fix with default paramters

– Bill

There’s room here for a joke about the tool “cleaning up” a Jimmy Hendricks fuzz guitar – and taking all the fuzz out.

The clip fix tools can do amazing things with a performance that overloads very briefly once every ten minutes or so. It can be the difference between having a clean show and not. What it will not do is rescue a show that lives in clipping distortion from the first note. There isn’t enough “music” there for the tool to use as a reference to fix the garbage. In that case sometimes you can convert to different garbage.

It joins a long list of tools that only work well before you realize you need them.

Koz

Actually, ClipFix doesn’t rely on surrounding “clean” audio to do it’s work. It assumes that anything above the Threshold is a clipped wave, and attempts to reconstruct that wave using a close approximation of a sine wave (the “cubic spline interpolation”).

The “Threshold” control is a bit tricky. Note that the instructions recommend amplifying by -10 dB before starting. ClipFix appears to scan the entire selection finding the maximum amplitude. The Threshold is a percentage of that amplitude, and is the point where ClipFix determines the slope of the “unclipped” wave.

– Bill

Right. But if you have enough clipping, you don’t have music any more, you have “Suite for Cubic Spline Interpolation and Variations.” You can ride over distortions here and there with that – it takes the crunchy out of the overload, but more than occasionally, you’re listening to the tool, not the performance.

It does do some interesting things with peak distortion that isn’t clipping and, as many have noted, you can apply it more than once and it sometimes gets better. Early on we talked about an auto-reapplication tool.

Koz

:slight_smile: :smiley: :laughing: ROFL

interesting

i can see how it makes it less unpleasant to hear when the D/A circuits recreate the original
but
unless my textbooks are wrong
it cannot come close to recreating what was there before the clipping – just make it less harsh

<<>>

Also right. Which is why it can “hide” peak distortions, cracking, or crunching every ten minutes or so, but is increasingly helpless in the face of more and more distortion and damage.

We still people they can’t recover from overload distortion because by the time they post for help, they can’t.

Koz