Improving a (faint) recording of a talk from original

Greetings all,

I have a poor recording of a public talk. I want to improve the full recording as much as possible. I’m using Audacity 2.1.0 under Windows 7.

I have attached a short WAV clip of the original, unprocessed recording. If anyone can do anything with it, and kindly let me know the steps you take in Audacity, I would be extremely grateful. I’ve posted this as a new thread to avoid confusion with another clip I posted that was deemed useless because it had already been processed.

Thank you for any help!

I’ve amplified 40 dB and ran Noise reduction with standard settings: -12 dB, 6, 1.

It sounds a bit better.

I can’t make out if this is real rain or bad noise because of a low recording level, cause I’m listening on laptop speakers.

But your recording seems salvageable.

The main problem is the “other” voice (“Yes, I do”). That voice is so low down in the noise floor that there is no way to separate it from the noise without badly distorting the voice. Similar to cyrano, I would amplify by about 40 dB, then use Noise Reduction. Don’t expect to be able to eliminate the noise - attempting to do so will badly affect the voices. The best that you can hope for is to reduce the noise without doing too much damage to the voices. I would suggest raising the “smoothing” setting to 6 so that the distant voice is a bit less “bubbly”, so the settings are: -12, 6, 6.

Thanks, Cyrano And Steve,

Using 12 - 6 - 6 on the noise limiter gives me a better result that I was able to achieve before. I wouldn’t mind sacrificing the audience comment, “Yes; I do.” if it would enable a better result.

Can anyone also advise on getting rid of the harsh, hissy S’s elsewhere on the recording. I have attached a short section of the original unprocessed recording with an example. Notice how harsh the 's’s are in the word ‘system’ (especially after amplification is applied). I can modify all of the offending S’s individually, using the EQ, by moving two of the sliders (around 4000Hz) down about 3/4 of the way to the bottom for the duration of each offending ‘S’. However, there are hundreds of these harsh, whistling S’s throughout the recording.

What about applying other mods to the entire recording? I heard that some people recommend compression and normalization (and some EQ to compensate for poor microphones). Would any such mods improve the overall quality to the average ear?



Try the famous Paul’s de-esser Nyquist plugin:

The download is in the first post of that thread.

How to install Nyquist plugins:

Try the famous Paul’s de-esser Nyquist plugin:


Thank you; that looks promising. However I am having some issue with this plugin download. Audacity 2.1.0 & Windows 7. After downloading the plugin from the link in the first post oWhen I try and move the downloaded plugin into my Audacity plugins folder, it takes forever, and I end up having to end the process in Windows Task Manager, otherwise, I can’t shut down my PC! …Not sure what’s going on there! Any clues?


I deleted your duplicate post about this, but added the Audacity and Windows version details from there.

Perhaps there is a locking handle on the file, for example your web browser or download manager. If so, restart the computer.

Also be aware that you must be logged in as administrator to add files to the Audacity “Plug-Ins” folder without prompts or permissions issues.


I have no idea, really…

Are you sure the download finished?

Did you try moving after a restart, or to a different directory (just as a test)?