OK, count me among those who feel completely ham-handed with Audacity’s newer, fancier versions of the NR tool…
I first started using Audacity in 2005, and really — for the most common types of annoyances I’d encounter, the old NR algorithm was sufficient to clean up the track. Just to be safe I’d keep a “dry” version of the track, and play around with the gain between the two until a decent balance was achieved, with any residual hiss or jingle cancelling each other out. None of it was particularly labor-intensive.
For 2.0 (and a couple of the older beta’s) I did check the tutorial and have tried following the advice where you set the first two sliders to zero, to simulate the old algorithm, then slowly creep to higher numbers until the desired sound is achieved.
Just not getting the results I used to.
I have uploaded a sample of what “silence” sounds like on one of my dry tracks. Is there a different tool I should be doing to filter out this noise-profile?
Ditto, It’s why I still use a copy of 1.3.X with the old noise reduction.
A noise gate can reduce or eliminate hiss in regions which are supposed to be silent, you can get a free noise gate plugin for audacity here … Noise Gate Tool? - #3 by Trebor
Give me that again? You created a dry but noisy track and a corresponding corrected, but bubbly track and faded between them until you got a good show? Did you get that from printed instructions anywhere?
I did notice that you could benefit from DC Removal. I amplified your work to see what was there and the battery voltage was significant. Effect > Normalize > Remove DC. Do this before you start editing or effects. DC or battery voltage inside a clip or song can mess with a number of tools.
Then, you can amplify your noise selection before you take the profile and that should give the system a booster shot. I think that’s actually what the Sensitivity control does, but I’m not the expert.
The two removal systems are very different and you can’t, last time I looked, get from one to the other with adjustments.
I used to have a krib sheet of what the controls did, but I don’t think I can do that since the last version.
Both the noise gates and the newer Noise Reduction try to recognize valuable content and route the correction around it. This allows much better reduction between the words, but worse reduction during them.
A common complaint with the older one was it was impossible to adjust it down far enough. The slightest nudge of the correction control and you had too much. You got the impression that if you could fully adjust the bottom 10% of the slider, that would do it. That’s probably why the poster had to resort to dual-channel tricks to get it to work. I never got it to do anything valuable.
I got rid of most of your hash and noise by just eliminating the DC signal.
Koz
You’ll should to remove the DC offset, (as Kozikowski describes), before using a noise gate.
The noise on “to-be-removed.wav” is essentially broadband : it covers the entire audio spectrum …
So Audacity’s noise reduction, even if it works as advertised, on optimum settings, won’t have much of a beneficial effect : Audacity’s NR method can only work if the noise primarily affects part of the audio spectrum, not where the noise is (approximately) equal across the entire spectrum as in this case.
Give me that again? You created a dry but noisy track and a corresponding corrected, but bubbly track and faded between them until you got a good show? Did you get that from printed instructions anywhere?
Nope, just trial and error. Worked sufficiently at the time that I didn’t question the “why”.
I did notice that you could benefit from DC Removal.
Now THIS is a tool I haven’t experiemented with yet, will definitely give it a whirl!
(Thank goodness the forum has file-hosting so I could upload the .wav of the problem…describing it and having folks guess, could have lasted forever. Like making broken-car-noises over the phone at a mechanic. lol)
It sounds like we need a better tutorial.
Which instructions/tutorial were you trying to follow?
The only differences between the current Audacity 2.0 version and the old Audacity 1.3.x versions:
The new version has a “Sensitivity” slider. For identical behaviour to the old 1.3.x versions, leave this at 0 (default)
The new version has an option to “Remove” or “Isolate”. For identical behaviour to the old 1.3.x versions, leave this set at “Remove” (default)
The “Attack / decay time” slider works in the current version but was broken in old 1.3.x versions. For identical behaviour to the old 1.3.x versions set this to zero. Setting to zero is not recommended as the default setting of 0.15 will usually work better. The old 1.3.x version was “stuck” at zero and the slider had no affect.
For behaviour that is very similar to the ancient 1.2.6 version, set the Sensitivity slider to about 8 and the Attack/decay to 0. This will usually produce worse results than leaving the Sensitivity slider at 0 and the Attack/decay at 0.15, but similar results to the 1.2.6 version.
Please note that the Noise Gate plug-in is now on the Audacity wiki.
If you have bookmarks/favourites to the Noise Gate plug-in, please update them to point here: Missing features - Audacity Support
Regarding Noise removal. I agree with Koz and Trebor that the first thing that is required for that audio sample is to remove the DC offset.
That’s not quite true, though interestingly this is the one situation where the old 1.2.6 version tended to work better than the later 1.3.x version.
It’'s difficult to say for certain without also having a sample of the proper audio (music/speech or whatever), but I’d guess that the 2.0 version will probably give best results if the Sensitivity slider is raised to 12 and the Attack/decay set to 0.10 (similar to the old 1.2.6 version).
Noise reduction is always a balance between removing noise and not damaging the audio that you want to keep. Generally it will sound better to leave a little noise and not damage the remaining audio, than to remove more noise and cause noticeable damage.
The absolute level of noise in that sample, after DC offset correction, is reasonably low, so as long as the signal level for the “real” audio is at a good level (peak level around -6 dB or so) then you will probably get good results by reducing the noise level just a little with “Noise Removal”, then use the “Noise Gate” to reduce the noise during silences a little more.
