8
Replies
Your fundamental problem here is that the audio is clipping, and that's what causes the buzzes and glitches. And it's pretty bad; the de-clipper found 175 of them, just in that sample. What's worse is that even if you run the de-clipper (which is inevitably imperfect, because it has no reference for what should be happening at that instant), you're still left with some almighty glitches like this:
That occupies the entire audio spectrum, and nothing is really going to get rid of it. You might be able to reduce the effect of it by painting out the offending bits and leaving the speech, but even that's not going to be perfect and it's an almighty pain to do all of them.
The problem here is that when it was recorded, the mic gain was either too high, or the speaker was a lot too close to the mic - or both. Once you've overloaded the mic preamp, you're basically stuffed; it's unfixable by any realistic means, as the signal has reached the limit of what can be digitised, and stopped dead - instant distortion. What you need to do is record this at a significantly lower level which will leave you some room for adjustments and alterations. As it is, the audio's going to need at least some limiting before it's acceptable for production, and probably a bit more work too. Can you outline exactly how this recording is being made? Almost certainly we can come up with some improvements...
AdChoices