Improving audio on recorded lecture


I have a lecture that I would like to improve the audio of. It looks like the wrong microphone actually got recorded, so while the speaker was wearing a clip-on which worked fine in the lecture room, the audio in the recording seems to have been taken from the podium mic and sounds really awful.

I don’t really know much about audio processing and I was hoping that somebody might be up for the challenge of taking a listen to the attached file and coming up with some pointers on what I should try to do to clean this thing up.

I know about the basic stuff, noise removal and compression, and while that my attempts (I didn’t fiddle much with the parameters of the filters) improves it quite a bit I would like to know if anyone has ideas on how to do better. I’m especially annoyed by the “barrel sound”… it’s not exactly echo but I think you’ll know what I mean if you listen to the clip. There’s probably a well known technique for improving this since it’s quite common to hear this when the microphone is too far away.

Of course I’d be interested to hear how good you guys can make this clip sound and if you’d be willing to share your magic tricks I’d apply it to the whole track… and you could get a very interesting lecture about Alan Turing and his code breaking work in WWII as a reward :wink:

Any help greatly appreciated,
Stefan Freyr.

I well known problem: reverberation, some claim to be able to de-reverb, but I’ve never seen it done.

Room Echo (reverb) is one good way to kill your show. There’s no “key” or single thing to tell the software to attack. A room echo is the actor’s own voice coming back to the microphone multiple times. You can’t tell the software to remove the actor from himself.

I got limited success with the Noise Gate plugin.

But all it’s doing is suppressing noise between words, not the words themselves. Try Level -12, Threshold -27, Attack 100.

There’s a lot wrong with that recording.

There are a lot of clicks caused by “drop outs”. Here’s a close up of one of them. You can see that the waveform is discontinuous just to the right of the playback cursor due to there being a bit of data missing. This is not repairable - you can’t put back what is not there. There a host of reasons that could contribute to this problem ( )
There is quite a high level of background noise. This could be improved a little using the Audacity Noise Removal effect.

There is a lot of “reverberation” from the sound bouncing around the room due to a poor acoustic environment and the microphone being too far away from the person speaking. There are few tools that can deal with this problem. The only one I know of is the “deverberation” effect in “Postfish”. Unfortunately Postfish was a project from several years ago that never got beyond beta status and is only available as source code. At one time there was talk of possibly adding this effect to Audacity but sadly nothing came of that.

There is noticeable distortion which sounds like it may just be a very poor quality microphone. The Equalization effect can help to take some of the edge off the distortion.

The sound is quite muffled possibly due to a combination of poor acoustics, poor microphone and poor sound card. The deverberation effect and the Audacity Equalization effect can help to some extent…

Basically it’s beyond repair, but anyway, attached is a before / after apply the above procedures.