1

I'm trying to edit audio for a podcast. In the audio of my voice, at some point I randomly start hearing fuzzy background noise that is only there when I am not muted. I'm not sure what caused it, but it sounds like a microphone issue.

I am a SUPER NOVICE, so i am very out of my depth. I tried the noise reduction tool in Audacity to get rid of the background noise, which it did for all intents and purposes, but it also ruined the speech audio and made it quieter and kind of tin like and hollow.

I have NO CLUE how to fix this. Can anyone help?

3

2 Answers 2

3

As you've not uploaded a sample, I can't say for sure, but the first step should be trying to understand where exactly the noise is. To do this, you can use spectrogram view that shows your audio file as a sort of heatmap.

Below is a sample audio file to which I've added a noise tone; this is just to illustrate, your real noise will look a lot different and probably a lot fuzzier.

Unwanted noise example

At the side of the screen in spectrogram view is a scale that shows the frequencies. Once you've identified the noise signal, check where it falls on the frequency scale, e.g. the noise in this example lies between 300Hz and 500 Hz:

Screenshot showing Audacity frequency scale next to noisy signal on spectrogram

Then you can take steps to fix this. The trick is to reduce the volume of the noise selectively without affecting too much of the rest of the audio. The easiest, fastest and most brute-force method is to use an EQ curve effect. This can be found in the "Effect > EQ and Filters > Filter Curve EQ" menu item. Make sure your track is selected, then open up this effect and you'll see a window like this:

Filter Curve EQ window

The bottom of the graph represents frequencies. The left axis represents gain (think of it as volume). The green line represents the gain/volume of the audio as it is currently. Clicking on that line adds control points that you can drag up or down. For this I like to create 2 points on the line as the limits to what frequencies to affect (here I added them at 300 Hz and 500Hz), then add two more points that I drag down to coincide with roughly where the noise is on the graph. The end result looks something like a triangle, as shown below:

Filter Curve EQ solution

You can refine this further by moving the points around a bit while listening to the audio on loop. This should result in slightly muddy or tinny audio, but if you do it right it will reduce the noise without sounding heavily distorted. If you upload a sample, I'm sure someone can point you at a much better method.

3

If you have a podcast file consisting only of human voices, you owe it to your future self to check out some AI-powered audio tools for cleanup. I've done the “conventional” ways, including the excellently thoughtful advice that salmonlawyer took the care to share, and I can suggest you try Adobe Podcast Enhance. Chances are that it'll also make your overall audio sound better, too! It's free, just drag-and-drop your file and see what it can do for you.

1
  • 2
    Oooh, that is remarkably good. I threw a badly resonant & reverberant file at it from another question here & the result really isn't half bad at all. Salmonlawyer's sample from sound.stackexchange.com/questions/52020/… [Though I had separated it to dual mono first & only uploaded the left channel.]
    – Tetsujin
    Commented Feb 3, 2023 at 15:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.