Can anything be done with this horrible quality audio?

First, I am new to Audacity.

I got it to help with some oral histories, especially the 9 hour history of which the attachment is an excerpt?

I’m going with no - not easily. The fuzzy rule is to give up when the voice and the noise are roughly the same volume.

How was that shot? There is one voice in there which is perfectly clear and loud. Was this a home recording of a Zoom lecture?

There are Artificial Intelligence methods of “cleaning that up,” but they require specialized sound card hardware and they’re not cheap.

This sample is stiff Audacity Noise Reduction

Drag-Select a portion of Noise-Only. No voice. Effect > Noise Reduction > Profile.

Select a larger portion and Effect > Noise Reduction > 18, 6, 6 > OK.


Thanks for trying.

The recording was made by/for the Imperial War Museum. I don’t know how or when. The interviewee was in his 20s during WWII. On the tape he sounds relatively elderly, so maybe late last century, or even more recently. I have e-mailed the IWM to see if a transcript exists.

I agree.

This version uses filtered noise (mostly around 1 kHz) as the noise profile, and the Graphic EQ effect was applied prior to noise reduction to boost the main “speech frequencies”.

That’s pretty good! And it helps that I generally know what he’s talking about. Can you give a newbie detailed instructions on how to duplicate this for the entire 4.5 hours? (9 hours was based on the wrong tape length.)

This is the EQ that I applied first to bring up the level of mid-vocal frequencies:

Amplify the track to bring the peaks back down to normal range.

Then generate some white noise into a new track and apply the same EQ twice so that it has a lot of noise in the frequency range that we boosted.
The noise track will be massively loud and distorted, so amplify it down so that it is about the same level as the noise in the first track (about -18 dB).

Use the noise track to get a noise profile into Noise Reduction.
and apply Noise Reduction to the first track - I used: 48, 3, 3.

Tracks menu > Mix > Stereo to Mono

Normalize to -1 dB.