Test Drive: iZotope RX 3 Noise Reduction Software Redux

Article Index

by Steve Cunningham

One of my most prized audio tools is not a vintage microphone or a boutique preamp. It’s a bit of software, albeit a very complex bit of software, that comes courtesy of the boffins at iZotope. We took a gander at iZotope’s RX plug-in back in February 2008, when it was at version 1 (as in Uno!). Since then others have released plug-ins that perform many of the same noise-reducing tasks, including most of the heavy-hitter software companies like Waves, Sony and even CEDAR, who is considered the granddaddy of noise reduction. Despite both product and pricing competition, iZotope’s RX product has continued to dominate the market, gaining a reputation amongst audio engineers as the go-to noise killer. It has certainly been mine for the past five or so years.

Not content to sit on laurels no matter how well-deserved, iZotope has tweaked and polished RX up to and including its just-released version 3 (actually released five days ago as of this writing). There are enough changes and improvement in the product that it deserves a second look, so let’s git ‘er done. Note that where I mention the RX product, you can assume that I’m referring to version 3.

iZotope RX


iZotope’s RX is a software Swiss Army Knife containing all the tools required to clean up just about any problem in audio. These include de-noising, de-clipping, click and crackle removal, hum removal, and spectral repair which allows one to target a specific range of frequencies for reduction or removal. In addition, RX includes a six band parametric equalizer, gain, channel and phase changing operations, plus sample rate conversion and dither. I think that about covers it, don’t you?

RX comes as both a standalone application, and as a plug-in for use in a stereo or multitrack software editing program. The plug-ins come in any of the usual formats, including AAX (for Pro Tools 11), RTAS/AudioSuite (for Pro Tools from version 7.4 through version 10), VST and VST 3, and Audio Units. RX runs under Windows from XP through Windows 8, and on the Mac OSX from 10.6.8 up to and including 10.8.4 of Mountain Lion, the latest version I have. It is also available as both a 32- and 64-bit application, which means that if your computer and OS can handle 64 bits, then with the 64-bit version you can access well in excess of 4 GB of memory. This is not only a benefit for working with long-form material, but it also means more processing can done in RAM where it’s faster than in previous versions. Finally, the software is written to fully support multicore processing, so if you’re fortunate enough to have multiple dual core or quad core processors in your computer, you’ll find that RX version 3 screams.

RX is copy-protected, and you have a choice when you register the product of using a challenge/response serial number system, or of using a $50 iLok (of course if you already have one you can simply add the auth to that iLok). I chose the iLok, and I’m doing that more often than ever because I find the iLok to be convenient when working on multiple editors and multiple computers. I can move the iLok between Macs and Win boxes without having to think much, and my plugs are available wherever I’m working. In either case, the program will work as advertised during a 10-day demo period, except that saving is disabled, so you’ll have time to see if it works for you.


As with previous versions, the standalone application is dominated by a spectral display. This shows both the waveform of the sound you’re working on, along with a representation in shades of white to orange of the frequency spectrum intensity at any location (the colors can be reset to your liking, if at all seems too Halloween-like for you). There’s a waveform overview above the spectral display, and as you play the sound a new Seek handle will walk its way across the overview. In version 3 you can grab that Seek handle and scrub your audio; that’s a new feature that has been missing for a long time.

At the very top are tabs corresponding to each of the sounds you have loaded, and it should be noted that this is another first within version 3. Previous versions only allowed you to load one audio clip at a time in the app. Having the ability to copy and paste between audio files as your making noise corrections within the app is significant. Equally significant is the ability to then process multiple individual files using a process that you set up, and knowing that each individual file will be processed in precisely the same way without having to save a processor preset and recall it for each file.

Along the bottom of the spectral display is a full toolbar of icons for identifying and selecting various regions in your sound file. A number of selector cursors are available, depending on whether you want to select a block of contiguous audio or use a brush tool to select small, freeform and/or discontiguous patches of sound. That section of the icon toolbar actually looks more like it came from Photoshop, with a magic wand tool, a paintbrush tool, and a lasso tool joining the previous selectors which tended toward selecting square regions.

At the very bottom of the window is a full transport, along with various information boxes that tell you where your cursor is at the moment, the duration of selected region, and so forth.

Just to the right of the display, a full complement of meters measuring amplitude in dB and frequency in Hertz helps get a handle on what’s going on in the frequency and amplitude domains of your sound. Both domains can be calibrated to show you a particular segment of either the frequency or amplitude spectrum in a magnified view.

On the far right of the window is a redesigned function selector interface, which is far simpler to use than the previous versions. Here, in a collapse-able module list, you will see all the modules available for noise reduction processing. Clicking any one of these will generally bring up a small window containing the necessary controls to set up that specific type of processing. Again, it’s a vast improvement and simplifies what is in the end geeky list of processors into something that’s not hard to use.

Best yet, there are tooltips that appear all over the place. And they’re not short three word phrases that give you only a hint about what a particular slider will do. The tooltips tend to come up in full sentences, such as “set whether the zoom slider controls the waveform amplitude scale or the spectrograph frequency scale.” That’s clear, isn’t it?

One of the standout new features of RX version 3 regular is that you can now add either VST or AU plug-ins into the application window (this feature was previously available only in the Advanced version). This of course means that when you render to a new file from the app, you not only have the benefit of the noise processing but also whatever effects processes you’ve chosen to add to the mix.


There really are no duds in the RX pile. But a few do stand out, and have been greatly improved.

For example, the de-noising processor has always been, in my opinion, one of the best available. Used intelligently and gingerly, it will reduce or eliminate all sorts of room noise or AC noise, and even works well on wind noise. The trick to getting good results is to go easy on the reduction parameter, and do as few passes as possible. It is often the case that simply reducing noise is more than sufficient, while trying to totally eliminate it may leave you with artifacts and less than perfect results. The improvement in version 3 is the ability to select multiple noise areas in one file, using the brush tools. So instead of having to pick one representative noise sample and teach the plug-in that it is the noise you want to reduce, or making multiple noise reduction passes on the same file to target different nasties, you can now select multiple noises in discontiguous areas and eliminate them all in one go.

Another good one is the de-clip module. Mitigating clipping distortion in a digital recording is one of the hardest things to do without damaging the overall sound. Simply cutting the high-end EQ will reduce the “fuzz”, but will leave the overall sound dull. And my experience with de-clipping plug-ins has been that they leave the file sounding much worse than it did with a little bit of clipping it. Frankly, the only effective way of eliminating clipping is to go through the file with a pencil tool, and manually round off all the square tops, a time-consuming task. The de-clip module in RX is one of the only ones I consider to be effective. It still requires a light touch, but the results have been more than acceptable. Version 3 automatically smoothes those clipped peaks in a manner that is usually inaudible, and will even reliably select a suitable threshold where the clip processing should kick in. It’s a good one.

The spectral repair module has also rocked the world of more than one of my clients. The ability to isolate very narrow frequencies over long durations and reduce their level, without affecting the rest of the frequencies, has bailed out many an actor. Do be aware that spectral repair is not always real time task, depending on your editor, so that plug doesn’t always work in all editors. Don’t worry, RX gives you a friendly warning and asks you to try another editor.

Did I mention that you can now save your work in a format called RX Documents, which saves a list of processes and settings performed, unlimited undos, and the original audio in one file? This is not a file for you to give to your client, but it is one to keep for yourself so if you have to go back in the future and undo it or redo it, you can do so by just pulling up one file. Brilliant.

As mentioned, all of these processors are available as individual plug-ins for use in your audio editor program. As you might expect, calling up any one of them brings up the same small window that you will see overlaid when you select a processor from the standalone application. The features, functions, and controls in the plug-ins are identical to those that come up in the standalone app.


As you might have noticed, iZotope’s RX is available in two separate versions; RX 3 and RX 3 Advanced. RX 3 is the regular version, while RX 3 Advanced has all the functions and features of the regular version, but adds many more parameter controls for more precise operation, and to accomplish tasks particular to a specific job like working with dialogue or audio forensics. Understand from the beginning that the regular version will have 95% of the functions the average producer or voice actor will need to clean up their tracks. Given the large price difference between the two versions ($350 versus $1,200 US retail), you might be better served by renting some studio time at a facility that owns the Advanced version, and has an operator who’s familiar with using it.

But if you do have the necessary resources (read: cash), the Advanced version adds a whole lot of muscle. My personal favorite is the de-reverb, which not only reduces or removes reverb processing that has been applied to an audio file, but can also simply reduce or remove room tone from a file. The results are not always perfect and depend upon what sort of reverb and how much was applied, but it’s the first plug I’ve ever used that does a credible job of removing room reverberation.

And the dialogue de-noiser does its job in real time. That’s right, it removes the common room noises, hum sounds, and other gremlins as the track is being recorded. There’s still more items in the Advanced, and it is not to be sneezed at. But it is expensive.


So what do I think of the new version of iZotope’s RX? I like it every bit as much as previous versions. iZotope has actually made some useful changes and additions to it. Opening multiple files at once, being able to scrub audio, and being able to perform phase rotation on voice tracks that have been recorded with asymmetric waveforms... these things are not eye candy. These are useful functions, and by the way they are included in the regular version of RX. It doesn’t come cheap, but if it can save your bacon on an important audition, I think it’s worth serious consideration.

Yes I know, all of the screenshots in this article include a big trial banner somewhere on the screen. Of course I’m running it in trial mode. I own a copy of RX Advanced version 1, which as I mentioned above I’ve been using for quite a while. The upgrade to version 3 is at the time of this writing about 400 bucks. That’s a fair chunk of change.

But given how much I like this product, how useful it’s been to me over the years, and quite frankly how much money I’ve made fixing other people’s audio problems with it, I figure I’ll make that money back long before version 4 comes out. So yes, I’m buying the upgrade.

iZotope’s RX (regular) is available at a US suggested retail of $349. RX Advanced is available at a US suggested retail of $1,199. Reduced upgrade prices are available on the company’s website store. For more information, visit www.izotope.com.

Comments (0)

There are no comments posted here yet

Leave your comments

  1. Posting comment as a guest. Your post will be moderated. Your email address will not be shown or linked. (If you have an account, log in for real time posting and other options.)
0 Characters
Attachments (0 / 3)
Share Your Location