Sometimes, I need to remove some high-pitched hissing from my audacity recordings. Like when you are exhaling to heavily while talking and into the mic. That can lead to these spiky sounds in konsonants like t, p, s and whatever else there is in your language.
You can not smooth this perfectly, but you can make it more tolerable, so that your audio is better to digest for headphone users. I tried listening to this with my DR77O Pro headphones. These are high fidelity headphones which I use for editing only, and they can be used without any additional equipment on your laptop. For real hifi you need some special gear actually, but this is already pretty good, especially if you think about how most people listen to audio: on lofi equipment anyway. So let’s optimize for that!
To get rid of high frequency noises, I recommend a low-pass equalizer curve, like depicted below. I personally don’t prefer hard cut-offs, that’s just a “physicist thing”, I guess. We prefer it smooth. I tell myself it yields a more natural output as well, but I don’t have proof for that ;-D
A systematic approach on how to remove hissing sounds from your audio in Audacity
You should try and find out what the evil frequency actually is, so you only cut off the absolute minimum. You don’t want any data loss for no reason, just get rid of the disturbing sounds. Those can go to zero amplitude, the rest can stay.
Here, I selected a short segment of the recording with a prominent hissing sound included. Just play around with this a bit and you will find what the frequency is you want to eliminiate. Then cut off before that with a steep curve. Only apply this equalizer curve to a small selection at first, for faster processing, to get a useful preview.
You can also just copy the section you want to play with into another track below your original recording, and then edit it there, so you don’t accidentally destroy your original recording. Make sure to de-noise and apply click-removal before. Otherwise these frequencies will show up in the spectral plot too, and that makes it harder to determine what’s actually the annoying high-pitched part of your voice data.
Quick and dirty, but it works
With this technique you can basically eyeball your spectrum. There are of course better ways to do this, but for the purposes of YouTube, this quick and dirty solution is sufficient. After all, the truth is:
- If you want better audio, get a better mic.
- If your audio is super awful, record new audio.
- If you can’t record new audio, make a voiceover.
If you’re a beginner editor you will probably be like »I am sure there is a fix, if you just have the right software, bla bla bla.« There is no fix for missing data. Trust me, if there was, there wouldn’t be so much sh·tty audio out there. So rather than worrying about how to remove hissing sounds in Audacity, you should focus on how to avoid them in the first place. Learn how your recording process actually works and what sources of disturbance are in your environment. Fixing something in place is always superior compared to fixing it in post.
Removing annoying harmonics in Audacity
The equalizer in combination with the spectrum tool is also quite neat to remove any multiples of some disturbing stationary frequency. One time, I was recording a video and didn’t have my Røde mic plugged in correctly. Or maybe my phone was creating this disturbance. However, I had this very annoying humming on the whole recording, and I could not filter it out with noise reduction (because it is not white noise and had quite a high amplitude).
So I made a custom equalizer spectrum to evict it and that worked not as bad as I thought it would. Sure, it will flatten your spectrum a bit and you will lose some clarity. But in my case, I really needed that footage and could not reproduce it. So that is how I partially saved the day. Here is what it looked like:
This one I also just eyeballed (yes, I am lazy. I do have f-analyzer tools, but it’s not like Google pays me a lot, so yeah, that was for another day). Anyway, you can estimate the frequencies pretty well because the spectrum analyzer snaps the cursor to the peaks automatically, and you just have to note them down. Pretty straight-forward.
Any questions, feel free to use the comment function. Check out my YouTube if you like, and happy editing!