Repair and restore audio
Learn best practices for fixing audio in Audition using the Amplitude Statistics panel, spectral frequency display, adaptive noise reduction, Diagnostics panel, and DeClipper and DeHummer effects.
Published
>> Adobe audition CC offers multiple types of effects and filters to repair and restore your audio but knowing how to apply these effects and filters in what order and really how to get the best sound possible involves just knowing the order of operations.
And the key here, there are three main things: [Analyze Normalize Process] Analyze, normalize, and then process.
[Analyze Normalize Process] Now the first step is trying to acquire the highest fidelity audio master possible.
That's where you want to start.
[Adobe Audition CC] And you can see if you look here [Adobe Audition CC] that all the files that I'm working with are in fact wave files [Adobe Audition CC Sample Rate] in a combination of different sample rates, channel configuration and bit depths.
Once I've located my uncompressed master file, now I can really begin this three step process.
And the first part of that is analyzing the statistics of the audio file [Amplitude Statistics] using the amplitude statistics panel.
[Amplitude Statistics] This is going to tell me everything I need to know including things like peak amplitude, [Amplitude Statistics] how much headroom do I have left.
So this is good, we've got about 7DB of headroom, that's nice because ultimately we're probably going to normalize this to also reveal more of that noise in the file.
[Amplitude Statistics] Clip samples, again, this is essential here, we don't have any clip samples on this file, [Amplitude Statistics] we're good to go.
[Amplitude Statistics] And then the actual measured [Amplitude Statistics] or recorded bit depth of this audio file in question [Amplitude Statistics] which was recorded 16 bit.
[Amplitude Statistics] Although, audition always processes natively at 32 bit.
Analyze first, normalize next.
Now again, we've got some background noise here.
You're going to have a hard time hearing that.
So I will typically take my dialog files, go up to effects, amplitude, normalize, [Effects Amplitude and Compression Normalize (process)...] and normalize them all to around minus three.
We're not compressing or limiting, we're simply globally, uniformly amplifying here just to reveal more of that noise and more of that raw content.
We can take a quick listen to this.
>> "When I think of play, I think of my childhood, and when I think of my childhood, I think of like..."
>> Okay.
And you can definitely hear there's some ambient noise and we can actually see it right there.
Now while we could in fact, highlight the noise like this, [Capture Noise Print] right click or control click and capture a noise print, [Capture Noise Print] that noise might not exist in all frequencies.
In fact what if it's just low frequency noise or maybe it's high end hiss or kind of wispy noise?
Well, we wouldn't want to just select everything because that's selecting all frequencies all the time.
[Show Spectral Frequency Display Shift-D] So this is where now we can use something called the spectral frequency display [0.00.741] to make a frequency specific selection.
[Capture Noise Print] Capture a noise print and now we could go into effects, noise reduction, process, [Capture Noise Print] and essentially use all of the tools here, all the various sliders and parameters [Close] to effectively attenuate that noise on our file.
Now if you don't want to take a sample, what if that's great if your noise is consistent?
But what if the noise changes over time?
Well, if the noise is changing over time, for that you probably want to use [Noise Reduction/Restoration Adaptive Noise Reduction] the adaptive noise reduction, and it does just that.
[Custom] It actually adapts to changes in noise over time.
[Custom] You've got a couple of simple sliders here to adjust to the amount of reduction, [Custom] the noisiness of the file, fine tuning the actual noise floor.
[Custom Close Apply] And this is a great filter again, if you've got changing noise over time, [Custom Close Apply] whether it's dialogue or ambient sound or anything similar to that.
Now if we take a look at this file here, once again I'm going to perform my amplitude statistics.
This is also dialogue.
Now this one we didn't have any problems with background noise.
We don't need to normalize this one because it's already normalized to around minus one.
It doesn't show any clipped samples and it shows me that we're at 32 bit.
But when I played this back I actually heard that there were some distortions on here, particularly in this little section here where he says, "On the couch."
>> "On the couch.
On the couch."
>> So this is an actual distortion.
Now it doesn't show me clipped samples but what happened here was this was distorted going into the preamp into audition, it digitally clipped the preamp.
So for that I can use our diagnostics panel where we have an effect called the declipper.
[DeClicker] Now this one is truly amazing because once again [Restore Lightly Clipped] I can choose a preset like restore lightly clipped, [Restore Lightly Clipped Scan] I can scan the audio file.
[Restore Lightly Clipped Scan] It detects 91 problems here.
[Repair All] I can choose repair all, and in a matter of seconds here, not only does it repair those clipped attack transients, but now it also removed that distortion.
So take a quick listen.
>> "On the couch.
On the couch."
>> So here's the before.
>> "On the couch."
>> Let's remove that.
Notice, now the waveform is been repaired and the distortion is gone.
>> "On the couch."
>> Now this file also had another problem.
When I went to the special frequency display, again, no background noise.
But up in the 19 to 21 K range there was something going on here.
So again, visually I could identify where that noise was.
And if I were going to add compression or limiting later, which I most likely would, I don't want to keep this stuff there.
I could take a frequency specific selection, just in this range, capture a noise print and then very effectively remove that noise in one simple step.
So the spectral frequency display is ideal for identifying different types of noise, where noise lives, and then again, can also determine what kind of tool ultimately you might use.
Now the last example that I'll show you here is actually based around an electric bass file.
And when I went to the spectral view this time around I could actually see very clearly that there were some 30 cycle hum present in this file.
And while I could use noise reduction to remove that hum, and of its subsequent harmonics, and I'll scan amplitude statistics again here.
We actually have an even better tool to do that.
[Noise Reduction/Restoration DeHummer] Under noise reduction, if I know it's a 60 cycle or 50 cycle hum, [Noise Reduction/Restoration DeHummer] we even have something called the DeHummer.
All right.
[Remove 60 Hz and Harmonics] Which again has presets to allow you to effectively remove those problem areas [0.00.752] very effectively, and again, keep on working.
[0.13.019] Now while there are no absolutes for audio restoration, [0.13.019] just using the simple tools and techniques of analyzing, normalizing, [0.13.019] and then applying the right noise reduction tool for the job will ultimately, [0.13.019] potentially give you the best results possible using the great restoration tools [0.13.019] in Adobe audition CC.
Contributor
Image by
Jonathan Petersen
