Weeding out the peaks

[theseus75]

I have a long series of tracks (an audiobook as some of you know), and there are inevitable outliers: points where the peak amplitude reaches almost to 0Db, while the rest hovers around - 9 Db.
I'm wondering if there's a way to soften these peaks, and put them down to a certain level or by a certain level so that - when I apply normalize - these outliers are keeping the rest of the tracks from having adaquate amplification.

It would need to be something that wouldn't effect anything else, though, so - say - something that put anything above - 6Db down to - 6Db but left anything that was already below - 6Db alone.

Is there such a feature in Audacity and I'm missing it?

Thanks for any help.

[billw58]

Two effects that might do the job are the Compressor and (if it was included with your installation of Audacity) SC4 (mono).

Try the effects on a short representative portion of the track until you find a setting you like.

For the Compressor, try Threshold = -6 dB, Noise Floor = -80 dB, Ratio = 6:1, Attack time = 0.1 secs, Decay time = 1.0 secs. "Make up gain ..." unchecked, and "Compress based on Peaks" checked. The problem with the Compressor is that the release time cannot be set lower than 1 second.

For the SC4 limiter effect, try RMS/Peak = 1 (peak detection), Attack time and Release time to minimum, Threshold = -12 dB, Ratio = 6, Knee radius = 1, and Make-up gain = 0.

With the above effects, the lower the Threshold, the more it will push down the peaks. Too low a threshold will start to affect the quieter portions that you don't want to touch.

Since you appear to be on a Mac, perhaps the best shot is the Apple AUPeakLimiter. Start with the attack and release times at .002 seconds and the "pre-gain" at 6 dB. The AUPeakLimiter is a bit different with its settings. It attempts to limit peaks to 0 dB after applying the specified pre-gain. The result is that you end up normalizing at the same time as you are doing the peak limiting.

-- Bill

[kozikowski]

I know it's fashionable to beat a sound track to a bloody pulp, but normal voice presentations -- before processing -- have 10 dB to 12 dB variation between soft words and emphasized ones. And that's not counting voice performers.

You would probably do well to capture the voice performance so no peak ever hits 0 and then apply Chris's Compressor at the default settings.

Chris's Compressor
http://pdf23ds.net/software/dynamic-compressor/

Chris does WAV evening-out and loudness management seemingly with no quality damage -- or damage that sounds very good. Those two ideas go a long way toward simple, effective voice performances with no tool tuning or custom settings.

Koz

[kozikowski]

If whatever you're doing is a big hit, you will be doing it for months or years and custom settings get tired in a big hurry.

Koz

[theseus75]

@bill - Thank you for all the settings; I will definitely try the PeakLimiter out if it does what I think you're saying it does. Much like pulling out any trace of sound from around my DAW, I'm trying to work on getting it in as best as possible to minimize after-effects. EG: Noise remover has been great and all, but I'd rather not use it at all.

@Koz - I'm liking Chris' Compressor, though it confines at least that phase of editing to the PC. In a very short section that I'd set back to a lower level for one reader (who's louder) and forgot to turn back for the second, softer reader, I used CC to boost it to the level of the other sections that were recorded far closer to 0db and the results were good.

Thanks guys.

[billw58]

I'm liking Chris' Compressor, though it confines at least that phase of editing to the PC.

Chris' Compressor should work as an Audacity plug-in on Mac just fine. It does on my Mac. Are you sure you downloaded the right version?

-- Bill

[theseus75]

@bill - Yeah, when I installed 1.3.12 Beta Unicode version, I read on the site how best to install to ensure that the plug-ins folder is kept with your app all nice and neat, and then - putting the .ny file in there - it's working now. Silly me.