3

Using Audacity, I can highlight a section of an audio file, select Effects > Noise Reduction > Get Noise Profile, then Select All, and then again Effects > Noise Reduction, and this time "OK" to greatly reduce the volume of the sound captured via that first Effects > Noise Reduction > Get Noise Profile.

However, I have to do this several times throughout a recording for various sounds which I want to lower in volume. Specifically, these are "inhalations" that are recorded when I am reading.

Is there a way to eliminate all sounds below a certain volume threshold, so that the sound of breathing is always completely eliminated, as well as all other faint sounds?

IOW, make it so that only the "speech" is recorded, and none of the "ancillary" sounds?

I would especially like to be able to add this to a chain, so that the entire recording is "scanned" and low-volume sounds converted to silence. Specifically, I would like to add it to this chain:

enter image description here

UPDATE

I downloaded the Noise Gate, but don't know where to go from there to install it into the Effects menu. After Effect > Nyquist Prompt... > Load, and then selecting the Noisegate.ny file, I see no "Noise Gate" on the Effects menu.

I can subsequently (after recording some sound or noise) select Effect > Nyquist Prompt... and see this, though:

enter image description here

If I select the "OK" button, I see this:

enter image description here

...but when I select "OK" there, no change is made to the recording, but I get a dialog, "Nyquist returned the value 250.000000"

9
  • 1
    There are plugins specifically designed to do this type of noise reduction, but neither example I can think of is cheap. Waves DeBreath or Izotope RX would be the 'go to' plugs for that task. There may be freeware equivalents, but I've never looked. A generic Noise Gate will do a similar task, but with less precision.
    – Tetsujin
    Commented May 7, 2017 at 9:10
  • Is "a generic Noise Gate" included in Audacity? I don't see it/one... Commented May 7, 2017 at 9:14
  • 1
    I've never used it, but most likely. It's a fairly basic effect type. If it has a search function, try 'gate'
    – Tetsujin
    Commented May 7, 2017 at 10:37
  • 1
    You have to download it: wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Noise_Gate Commented May 8, 2017 at 15:19
  • 1
    @SimonBosley: Please see my update. Commented May 8, 2017 at 16:50

3 Answers 3

2

I doubt you will get an acceptable result with a simple gate setup: while you may get rid of low volume breathing sounds, it will also chop away natural endings.

Noise reduction plugins with noise profile capture functionality tend to distort the audible material too - making it sound cheap and wobbly.

I can strongly recommend the DeBreath plugin by Waves:

It works very well!

I am not affiliated with Waves in any way, and I'd like to give you other alternatives if I just knew about them.

You may be able to achieve something similar with a sidechained gate/expander, where the key signal is high passed to expose as much unique breath sound as possible, but you wil most likely run into the same issues as mentioned above (just to a lesser extent).

Update 2022: Alternative options may be iZotope RX or Waves Clarity VX / VX Pro, in particular if the unwanted noise is background noises like traffic, motors or fans. Waves Clarity is the newest and it delivers a very impressive result:

5
  • ...does it work with Audacity? Commented May 10, 2017 at 14:26
  • Yep, standard VST... Commented May 10, 2017 at 14:59
  • Strange that their help doc makes no mention of Audacity. Commented May 11, 2017 at 13:10
  • "while you may get rid of low volume breathing sounds, it will also chop away natural endings" I respectfully disagree... If used correctly, a targeted gate can be undetectable. They have been used effectively for decades! DeBreath is a good tool though, but you say you don't like "noise profile capture functionality", but what do you think DeBreath is? It's basically a gated noise capture algorithm. Personally, I think any of the tools you mentioned can be used very effectively for their intended situations.
    – n00dles
    Commented May 12, 2017 at 12:27
  • @MarcW Agreed, if used correctly the gate approach may work (that is kind of implied in the sentence, right?). Using it "correctly" may imply advanced sidechains and even automation in this case - it just takes one case of a chopped ending/unwanted breath slipping through to support my point. As you retweak threshold for that case, new ones may get introduced elsewhere...your typical gate conundrum. DeBreath is much simpler and works well. I did not code that plugin, I do not know the underlying algorithm, but I don't see any capture profile functionality... not sure what you're going at? Commented May 15, 2017 at 6:02
4
+50

It really depends on the material; are these sounds that you want to remove truly peaking at a lower level than every bit of the speech? How about in loudness or 'energy'?

You need to look into the kind of sounds you want to remove and see how they differ from the ones you want to keep. This is key to being able to separate them.

Are they of a shorter duration than the sounds you want to keep? Are they at a different frequency? Do they have some sort of regular pattern?

I'm just trying to get you to think in the right way. You need to first define a difference before you can look for a tool to remove them.

You have established that the unwanted sounds are at a lower amplitude than the sounds you want to keep. This makes things very easy; all you have to do is find a tool that will allow a signal to pass through(or open the gate) if the amplitude passes a certain threshold, and reject any signals that fall short of this threshold(close the gate). It's almost as simple as that.

I say almost, because there are other parameters to consider. How fast do we want the gate to open and close? because we don't want any sharp clicks and we may want a smooth transition through the attack and decay of this opening and closing gate. Also, do we want the closed gate to completely attenuate the signal, infinitely, or do we want to just gently attenuate the signal by, say 12 dB? Do we want the left and right channels to be 'gated' individually, or do we want to treat them as one signal? These are the basic extras that a generic gate provides.

If you are having trouble with Audacity's Nyquist gate then I would suggest you look for a free VST gate online. There are many to choose from, pretty reputable companies release free VSTs if they are simple enough, like noise gates.

best free noise gate vstau plugins from bedroomproducersblog.com
unfiltered audio g8 cm free vst au noise-gate plugin from musicradar.com
Other cool free VST FX(Voxengo)
Immense list - free Gates/DeEssers

I think your main problem is understanding how to use these gates or de-essers(de-essers are just gates that target high frequency sounds like "sssss" sounds (or "sibilant frequencies"))

I explained the basic parameters of a gate, but there is one more that could be helpful to you, it is an internal sidechain feature, sometimes called other things but basically, it allows you, like the de-esser, to target certain frequencies with the gate. It usually consists of two parameters, an upper and a lower cutoff frequency. So if you set the lower cutoff to 2 kHz and the upper to 5 kHz, then the gate will only be triggered if the amplitude of this range(2-5 kHz) passes the threshold. So these inhalation sounds could effectively be targeted in this way.

So a word on setting up a basic gate;
To initially set up a gate, you want to set the attack/decay to a low level; zero if you like, set the attenuation(level reduction) as low as possible, then sweep the threshold down, starting at 0 dB. This will give you a good idea of when the gate triggers, so you can then set the threshold to the appropriate level(just above the unwanted noise), then adjust the other settings to fine tune.

Good luck!

1

Yes sortof maybe.
Special plug ins can help a lot depending on the exact problem.

The only correct thing to do is to get a better recording. Teach the speaker/vocalist how to use the mike to avoid breath sounds.

Use a quiet room to avoid ancillary sounds.

Noise reduction works quite well to remove hissy type noise but other sounds, NOT properly called noise in this context, need other methods to help eliminate them.

And yes you can remove all sound below a level with a gate. You might not improve the speech that remains though as much as you think it will.

1
  • Thanks; I am the speaker; I record Mark Twain material (on YouTube and my Twain facebook page). I cannot control sounds in the room. The only place I can do it is close (enough) to a refrigerator that I get the sound of its motor intruding. Commented Jan 26, 2020 at 13:43

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.