Clean old cassette audio

I have mp3 files of my grandparents’ life story, recorded originally onto audio cassettes 12 years ago before they died. The audio is pretty bad. 2 years ago I followed audio cleaning instructions for Audacity (sorry, I don’t have a URL for the instructions I followed) and cleaned one of the files. The result was far better, but included artifacts resembling some kind of digital wind chime. Below are links to the first 13 seconds of the recordings. The 1st one is original, the 2nd one after being cleaned by audacity:

original: http://www.zerogravpro.com/Personal/original.mp3
cleaned: http://www.zerogravpro.com/Personal/clean.mp3

I really need some tips on how to either (A) remove the artificats from the cleaned version or (B) re-clean the original so that artifacts aren’t created. Any help would be much appreciated.

Thanks!!
SweatCoder

(I didn’t listen to your files, 'cause I’m at work.)

:frowning: With noise reduction, sometimes “The cure is worse than the disease.” You can try a less aggressive setting.

You just have to experiment with noise reduction, noise gating, filtering and equalizing. Usually, you can make some improvement but not always. And, it’s rare that you can take a poor recording and makie it into a great recording.

For example, Even with professional-modern software, on-location movie dialog is re-recorded in the studio. It’s still important to start with a good recording.

If the audio is difficult to understand, you might consider typing-up a transcript to go along with the recording.

The original recording is at a very low level. If you record it louder, that may help. Aim for a peak recording level of about half the track height.

Other than that, it’s a trade-off between hiss and wind chimes. If you turn down the Noise Reduction amount, there will be some hiss left, but it will probably sound better overall than the jangling wind chimes. There is no way to remove the jangling wind chimes.

I processed the original MP3 as follows:

Import into Audacity
Split Stereo to Mono
Delete one track
(now I have just one mono track, which is all you need)
Select entire track
Effect > Amplify - accept defaults
Select a region of noise only and export as WAV
Export the entire track as WAV
Open DeNoise (by Brian Davies)
In DeNoise, Preferences, check “Pause on completion of processing”
Open the noise sample
Set Limit Noise Reduction to 12 dB
Click Start
When processing is complete, select the noise sample in the upper “waveform overview” bar
Click Sample Noise - this gets the “noise profile”
(DeNoise reports the Noise Floor as -25 dB - that’s really noisy!)
Click Resume, and Accept New Settings - processing is completed
Open the full recording
Click Start
When processing pauses at the end, click Resume

Import the resulting file into Audacity
Effect > Low Pass Filter, frequency 14000, rolloff 48dB
(Analyse > Plot Spectrum shows no frequencies above 14 kHz)
Effect > High Pass Filter, frequency 40, rolloff 24 dB
Export as MP3 VBR quality 3

I think it is quite listenable, and possibly the best you’re likely to get. I tried it again with 20 dB of noise reduction (in DeNoise) and there was some subtle “phasing” type sounds (but no wind chimes or birdies) and the background hiss does not sound much reduced from the 12 dB version.

– Bill

Using a similar method to Bill, but using Audacity’s Noise Removal with the modification described here: https://forum.audacityteam.org/t/noise-removal-tests/15869/1

That’s a definite improvement over the stock Audacity NR. Subtle, but better. Fewer artefacts, but the DeNoise version has virtually no artefacts. The upgraded Audacity NR seems to have a slightly brighter sound, despite the more aggressive HF rolloff. This is a comparison of the spectra of the first phrase (from about 0.5 seconds to 4.3 seconds). AudNR is Steve’s cleaned.flac, DeNoise is from my procedure.
FRCompare.jpg
– Bill

You can probably ignore the differences at the extreme high and low frequency ends as I probably filtered it more aggressively than you did (particularly the low frequencies). But, yes, less artefacts than the stock NR. Enough to warrant a 20% increase in processing time?

Yes.

– Bill

@SweatCoder - why are you working with MP3 files - these are compressed files designed f or use on portabel music machines - bad for editing/processing as some of the audio has already been removed in the MP3 compression process. Do you by any chance have uncompressed (e.g. WAV or AIFF) files? Or do tou still have the original tapes and decent tapedeck so that you can record tme into Audacity and work with them uncompressed there?

You can always produce compressed MP3 files for use and sending to family etc. - but since these are family archive you should also be saving and backing up the WAV files too, as well asw making MP3s.

WC

+1

Consider the material here - it is an irreplaceable family archive recording. The tapes are old and will not last forever, the recording decaying slowly over time anyway. In this application I expect that users would value the best-possible-result over time-taken-to-process.

WC

I like your noise removal better, the link points to a thread, could you point to the post where I can read the method you followed to remove the noise ?

thanks