Managing hiss on old spoken audio recording

I’m currently working on restoring audio from a 63-year-old 1/4" reel tape.

The reel player put a pleasnat 60-Hz hum on it, which Audacity removed easily enough… but I’m trying to improve what remains as best I can.

To this point, I’ve had mixed success with noise reduction, limiting, normalization and equalization. Admittedly, I’m not much of an expert in signal processing, but I’m not afraid of figuring things out, either. That said, the thing that seems to be most difficult to address is managing the noise without further degrading audio quality.

At this point, it seems like a really useful tool for me is one that doesn’t exist. Specifically, I’m interested in normalizing audio which exceeds a certian db level (or de-amplifying audio that doesn’t). The idea is simple - the background noise generally stays quite low, so it seems as though if I were to apply normalization / deamplification using a sort of amplitude filter / mask it might go a long way toward reducing background hiss…

With the notch filter?

Don’t bother with limiting, it won’t help. The dynamic range of tape is relatively low, so no need to reduce it further.

What speed does your tape play at? 7.5 inches per second? More / Less?

I had better luck creating a custom equalizer curve that picked up just past 60Hz. The notch filter worked as well, though.

The value of limiting is it allows me to normalize it to get better volume. The problem is, it brings up the hiss as well, so I was hoping I could maybe force it to ignore peaks below a certain amplitude, though I’m certain that won’t work as well as I’d like.

The original recording was at 1-7/8 IPS. I played it on an old reel-to-reel at 3-3/4". Best part is that it’s a mono tape with recordings on both sides, but the reel player was stereo - so I captured both sides simultaneously, one forward, the other backward, one in each channel.

So my processing workflow has been:

  1. Reduce speed on both channels by 1/2.
  2. Revrse one channel
  3. Apply equalization to cut out 60Hz hum
  4. Limit spikes to -6 db for better normalization
  5. Normalization to -0.1 db
  6. Gentle noise reduction to knock the hiss back a few decibels.

So I’m trying to find a better workflow that will help me amplify the voice without forcing me to rely so much on noise reduction to kill the hiss.

That sounds about right.
Considering that the original recording was at 1-7/8, there’s probably not much real audio content above about 10 kHz, or below 60 Hz, so in your equalization step, you could roll off those frequencies. If it’s a speech recording, you could roll off a bit more, say 100 Hz to 7 kHz. This should slightly improve the amount of hiss, but also help the noise reduction to do a better job.

You would do that with a “noise gate”. That can certainly help to reduce (or even eliminate) noise during “silences”, but the down side is that in parts where the level is above the threshold dB, the noise will still be there. This can cause a weird “breathing” sound as the noise floor comes and goes.

If you want to try a Noise Gate, you can get one here: Missing features - Audacity Support
Installation instructions for Linux are here:
I made my own voice impression of Mel Blanc on my smartphone including some voices editing in Audacity. Here’s original video.

luke11685, did you accidentally post in the wrong topic, or does your post relate to joelgraff’s question in some way?

Oops sorry. I thought that besides setting up old spoken audio recording you might have digital voices of some old voice actors for instance Mel Blanc or even Walter Disney. Is this thread about some sort of microphone voice reversed engineering? You know voice impressions of such old vintage voice acting skills if it will go further.