Noise Sample for Reduction

Hello

I’ve read that a longer Noise Sample will give better result for Noise Reduction.

  • Is it ok to make a track for the Noise Sample in which I would copy several noise samples taken from different places in my sound?
  • If yes, shall I merge them, or can I keep the splits (so I remember where they are)? Shall I care about the way they are connected?
  • If I need to delete a poc from the noise sample, do I need to care about how the 2 parts (before & after poc) will be connected together?

(the noise in my sound is constant. It comes from the psssshhhh from the sound system in the room where I was recording the conference. My recorder was too close from it. And the recording is very weak compared to the back noise)

That was true of the old “Noise Removal” effect, but less so for the new “Noise Reduction” effect.
More important for the new effect is that the noise is representative of the noise that you wish to remove.

In both effects, the “noise profile” is created through statistical analysis of the selected noise, but the method for doing this has changed between the two effects. The ne effect is able to obtain its statistical data from a shorter selection than the old effect.

There is a minimum length for the noise sample (the length depends on the sample rate). The effect tells you if the selection is too short.
Generally it seems to give no benefit to extend the noise sample by repetition.

If you don’t have a long enough section of “noise only” to make the noise profile, then you can create a new track (must be the same format as the track that you want to process), and copy/paste short sections of “noise only” from different parts of the original track into the new track, then use the new track for the noise profile.

That’s going to be tough. Don’t expect to be able to remove the noise completely - attempting to do so will make the remaining sound horrible. You will just have to try and strike a balance between “noise” and “not too much damage”.

Thanks

Yes it is hard long work since the recording has a weak level…
Doing my best. Trying so many different settings…

I noticed that removing some frequencies with Notch Filter before doing Noise Reduction does help a lot. Then Noise Reduction doesn’t need to be so strong and the sound is less damaged at the end.

Yes, that’s a good tip if the audio contains a constant hum and/or whistle type noise.

In the same way, applying a high pass filter (with the frequency set quite low) can be helpful for reducing low frequency rumble prior to Noise Reduction.

Yes, I did that too, it really worked.

I’m also wondering if Equalizer could help… After Noise Reduction, by trying to reduce the frequencies where some noise still remains. But it is really hard to find out what are these frequencies, even using the Plot Spectrum tool (it is much easier on a Noise Sample)

Try spectrogram view … Spectrogram View - Audacity Manual