Many people think of iZotope RX as the software most suitable for post-production or forensic applications. But this kind of audio restoration software also has its uses in mastering contexts. Sometimes you get a mix that requires a little restoration. With RX 7 out, we thought we’d cover the most common uses for this type of software in mastering.
But first, there’s the matter of how to implement RX. In my practice, I find it best not to work with RX as a series of plug-ins, but in standalone mode. I identify the issues early on in the process. Sometimes I take care of them before applying any processing, and the RX’d file becomes the basis for my master. Sometimes I wait until capturing the audio from my outboard hardware, but before the software limiting stages. It depends on the process and on the project. If files have come in at different sample rates in any given project, RX in standalone is also quite handy for sample-rate conversion.
Whatever I do, I’m always careful to create a new file for anything with RX processing; you never want to destroy an original copy of a mix, or the original file of your pre-rx’ed capture after hardware processing.
Mixes can often come in with clicks and pops, either from digital issues in the mixing engineers' setup, from mistakes in the recording process, or for mysterious reasons altogether. Sometimes clicks and pops make it past mastering into records that you and I could both name!
However, you can frequently handle clicks and pops discreetly using the De-click module. The key is to tune the frequency-skew to the range you perceive the click to occupy, and to adjust the widening so it’s only focusing on the click. Audition the results with the “output clicks only” box check ticked on; this way you’ll hear that horrid, errant click in all its detail.
If you’re also picking up some of the actual sounds around the click, you know you’re processing too heavily. Mess about with the sensitive slider until you’re picking up just enough. Uncheck the box and preview the results: it should be as though the click was never there.
Be sure, when processing, never to grab too much of the region around the afflicted area: you have to go in like a surgeon, and leave no trace behind.
I’m surprised how many tracks come my way where people leave in mouth noises, which makes RX’s Mouth De-click module a godsend. Similar controls to De-click are offered, and your operation is pretty much the same: mouth-clicks can be high pitched or sit lower down the frequency spectrum, depending on the vocalist’s throat and tongue position. Find the frequency range that works best, then tune the sensitivity to the least amount that gets the job done (the same goes for widening).
Only process regions specifically afflicted by mouth clicks. If there is a one second stretch between two stray month noises, don’t highlight the whole thing and click process; select each instance and fix it individually. This will yield for a far more transparent result, and won’t take away from audio that is perfectly good as it is.
When you’re done, do yourself a favor and hit undo to hear the original again. Then hit redo. Toggle back and forth a couple of times to make sure the only thing you’re missing are mouth noises.
Ideally mixing engineers will deal with plosives on their end, but not all engineers are that discerning. A sloppy mix-engineer might have not thought to address them (I’ve had a few such projects recently). Or, you could be mastering a live record, where all sorts of variables are at play that might make editing plosives unfeasible.
Whatever the case, if you have a module like RX’s De-plosive, you can reduce plosive issues with minimal effect to the overall sound, especially in an acoustic mix before drums come in to impact the low end.
This doesn’t mean, however, that you can just set-and-forget the plosive module and expect effortless results. Sometimes, for a proper job without artifacts, it makes more sense to look at the spectrogram, find the offender in the frequency spectrum, and from there, switch to a tool that grabs specific frequencies, such as the lasso or the paintbrush. Now you can highlight the most heavily shaded parts of the spectral graph during the plosive—for these are the biggest offenders—and run De-plosive in a frequency dependent manner, or Spectral Repair, if you think that might do a better job. More on Spectral Repair later.
Every once and a while you get a project that has audible distortion. It could be undesirable clipping, or it could be something else entirely. All you know, from your vantage point, is that the distortion is undesirable to you.
But it might not be undesirable to the band. Your first move is to call the point person on the production team and ask if that distortion is intended. If it isn’t, your next move is to contact the engineer, if you can, and negotiate a better mix. This option isn’t always on the table, so luckily software like RX has ways to fight this distortion—tools which can, to some extent, repair mangled material.
A lot of the time the De-clip module does the trick, even if the material isn’t clipping per se. The process is relatively simple: highlight the passage that is distorting, have the module suggest its processing, audition the results, and tinker to taste.
Repairing distortion is best done in specific, sporadic places, and not across passages of more than a couple of seconds—and that’s speaking liberally. You may have to spend a bit of time to get all the nasty bits, but it’s still worth it.
Sometimes, the distortion is more of a high-end crackle than a clip. Here, the De-crackle module, tuned to attenuate high-frequency distortions, can often work.
Sometimes acoustic material isn’t recorded in optimal spaces, and here, it might be best to de-noise a passage. Ideally you shouldn’t have to implement a de-noise module for too long a time—a solo acoustic introduction to a fully fleshed out song, for example, could be all you need. In general, try to air on the side of caution, because the ear is far more forgiving to the sound of noise than to noise-reduction’s usual artifacts. But give it a go if necessary.
When it comes to de-noising, it might be better to implement this process after EQ and compression, because these processes can change the quality of the noise. You may have to commit your digital processing before implementing noise reduction if you’re working entirely with software. If you’re using outboard gear, I find it’s best to denoise after the capture stage. Ultimately, the decision of when to de-noise is your call.
Recently a mix came my way with unflattering distortion on one guitar part in the left channel. It only occurred in one part of the tune, so I checked with the band, and yes, if I could do something about it, that would be great.
So I loaded the track in RX, found where it happened, and took out the distortions with the De-clip module—but only in the left channel, and only for a small duration of time with each pass.
This is what I find wonderful about audio restoration tools: the ability to fix something that exists solely in either the left or right channel, something that would necessitate a time-consuming creative solution in yesteryears, is very easy to fix now.
And it’s not just distortion: clicks, bits of unnecessary noise, and more can lurk in one channel or the other. In these cases, why would you process the whole mix when you can actually delve into the corners and excise the issue with a more surgical hand?
Mixes come to you in all sorts of states. Ideally, you should have a great relationship with the mixing engineer, be able to call that person on the phone and say, “hey, the balance of the bass is off,” or “did you mean to have the snare drifting towards left?”
But this discounts the unfortunate politics that go into record making. Say the mixing engineer was not only the producer of a track, but was involved in marketing the material. Say you respect that’s person’s ability in the producing and marketing arenas, but the mixing is badly balanced. To tell that person their mix is somehow inferior might be to shoot yourself in two different feet if you wanted to work with them again. You don’t want to aggravate a political situation and make your artist’s life harder.
Here we may have to do a measure of “rebalancing” ourselves. Most of the time this is not—repeat not—an audio restoration task; if the vocal doesn’t feel centered, and we need it to feel centered, simple panning or balancing tricks within the DAW can sometimes help us.
However, I do use audio restoration tools every once and a while to accomplish rebalancing. Perhaps the problem is relegated to a frequency band, in which case RX’s ability for frequency-specific processing across all its modules is a boon. Sometimes the Mixing module can redirect the energy in a more natural—and more nuanced—way than the panning tricks of a DAW.
Now, with RX 7, you’ll note the new Music Rebalance module. This is an interesting tool, and one that can do some good if used with a subtle hand (as is the case with most mastering tools). Say the all the drums are a bit too soft, or too buried; you could use the rebalancer in small degrees to raise the drums just so. Alternatively, you could isolate the drums drastically, bounce them, and export them as a parallel track into the session. However you would need to be very, very careful. The levels would need to be low (lower than parallel compression in a mix on even the subtle side), and, you would have to make doubly sure the smearing or artifacts trade-offs were worth it. These issues would have to be so unnoticeable as to not cause harm, while still justifying their use in the first place.
With spectral tools of the sort RX affords us, we can get aggressive with excising the cancer of unwanted noise. A page-turn in a solo classical guitar piece can be extirpated without noticeable artifacts. A digital click that occurs smack in the middle of a vocal’s frequency can be alleviated by spectral repair. Even a bad edit can be seen to and dispensed relatively easily.
The spectrogram allows us to see where the offender is in the time and frequency domains. The density of color in a frequency range shows us how strong the offender might be, and can help us pick out the anomaly.
When we’ve determined the troublemaker, we can highlight the spot with accuracy. Spectral repair tools can then analyze the signal abutting the problematic selection, and in turn, use that audio as an example for how to tailor the sound in its crosshairs.
Choose “Attenuate” for graceful mitigation of the sound, or, if you need the passage to fit into its surroundings, try “Replace.” I find Replace works best for short durations of audio, or else it gives the fake away too quickly. Replace came in handy just the other day: I had a master where, at the end of the tune, one could hear a clicky sound—not a digital blip, but almost as though someone had clicked off an old Walkman. On the spectrogram, it looked like this:
In the master, you couldn’t hear it.
Because you select a troublesome slice of audio however you like, you can also grab complex sounds like a siren, or an audience member hollering: the rise in pitch will be visible to you, and you’ll be able to trace around it and minimize the sound. It may take a few passes of a given process: be sure to keep things sounding natural.
You may be called upon to polish an old, analog transfer into something more clear. In these cases, keep in mind that the ear is far more forgiving to crackles and old-timey hiss than it is to the artifacts of noise reduction. Spectral De-noise might come in handy if you don’t pile it on too heavily, though a simple low-pass filter might be better. The one in RX is remarkably good for these sorts of things. De-crackle has been known to work too in mitigating the occasional crackle, and De-click can work on intermittent clicks.
The ends of tunes can often be repositories for sloppy behavior; in louder than ideal mixing environments, engineers might not hear what drops out as a tune ends. So maybe an amplifier’s region might cut out before the rest of the band fades to silence—maybe this is even an intentional choice, because something weird happened in their take before the end.
Recently I got a raucous rock number with this exact issue: the hiss from an amp died before everything else did. Often a bit of reverb brought up on another, multed track can gel endings in this circumstance, though one must be very judicious in making sure the reverb is appropriate for the tune, and is timed to perfection. However, on this track, Ambience Match was a more suitable job: it put the sound of the amp hiss back into the ending. Whatever tool you use to rescue the end of a tune, make sure you spend some time getting it just right.
Here’s the best piece of advice I can offer: don’t go overboard when it comes to audio restoration. Diving deep into the weeds, listening closely for such errors in audio, you’re going to start to hear things you’d love to take out. New clicks and pops will avail themselves; you may wind out taking more than you ever should, leaving only weird artifacts in the place of forgivable noises. You could very well drive yourself crazy. Learn to hear the master not just as an engineer would hear it, but as a casual listener. If it won’t bother most people, you may be better leaving it well alone! But if you think you can get it out with your trickery, give it a try. Just make sure you never destroy the original—that way, you can rest assured your experiments won’t mar things irrevocably.