Hands-On Review: iZotope RX 3 Audio Restoration Software







If you're a beginner or a professional engineer, you've most likely had this painful experience at one point or another: sitting there, tweaking EQs and filters, trying to salvage a busted clip of audio to make it usable for a project. What has brought you the misery this time? Is it wind and background noise polluting your dialog track from the field, or the harsh, bright reflections from your untreated garage making your vocal track ring excessively? Most of us start with high hopes and patience, then soon enough we embrace reality, bail on the effort, and get ready to re-record. In the real world, you don't always have that option.

Where other brands offer several separate software solutions for cleaning up individual problems in your tracks, iZotope presents the newest version of their solution that's a comprehensive Swiss Army knife for resolving several audio issues. RX 3 Advanced is a collection of modules with specific tools to operate on damaged audio and repair it at the root of its problem. Raising the bar in this new version, I was excited to go hands-on and see how they upped the ante. As a professional engineer working with post production as well as music, I used RX 2 Advanced as my "go-to" software to rescue compromised audio. Taking RX 3 Advanced for a test run, I found several enhancements that really make this tool faster to use and more accessible to general users than other related solutions on the market.

The range of modules inside RX 3 Advanced offers repair options for mulitple, notorious issues that plague engineers working with audio tracks. Added to the the RX collection of Declip, Declick & Decrackle, Remove Hum, Denoise, Spectral Repair, Deconstruct (balance noise versus tone), Gain, EQ, Channel Operations (stereo balance, stereo phase, Azimuth, or center versus side control), Time and Pitch, Resample, and Dither, is the new Dereverb tool. With one purchase, you are covered for several tasks, such as transferring audio from analog sources, mastering, noise reduction in many capacities, and even forensic work like trying to bring out one source in a mono or stereo track that's filled with noise. This audio-repair software can run as a stand-alone application, and supports 64-bit AAX (for Pro Tools 11), RTAS/AudioSuite VST, VST 3, and Audio Unit plug-in formats.

On the surface, RX 3 Advanced features a new UI and design, which not only looks cleaner, but also makes options easier to access and utilize. Moving on from the previous "block" look of the RX 2-generation software, modules on the right can now be viewed in a complete list without any scrolling, and can be collapsed as icons to maximize the real estate on the screen so you see a larger waveform. The tools on the bottom are also arranged neatly, making the interface even more compact than it had been. While some new features like record and input monitoring are available on the transport, significant updates to the previous version, such as session recovery (for crashes) and the update to the Undo History (where an RX Document can save undo history for future sessions) can dramatically improve workflow across several sessions.

Under the hood, RX 3 Advanced offers some true layout and operational enhancements. One thing you don't see directly is the new support for multicore processors. iZotope invested their efforts so that processes are spread more evenly across the cores, and the results are impressive in terms of saving time.

I opened RX 2 Advanced and RX 3 Advanced side-by-side to compare the same processing procedures on the same audio file. One great new innovation is that iZotope has included a time countdown in their progress bars, so I didn't require any timing tools to get a basic idea of how much faster workflow would be. The sample audio was a 6-minute 24-bit/48kHz stereo file, and Denoise was the module I chose to use for the comparison challenge. Since the RX 2 and RX 3 interfaces are somewhat different, I used similar settings to achieve the best-quality result; reduction at 12 dB with the threshold at 0 dB. RX 2 predicted a 3-minute, 20-second processing time, while RX 3 clocked only 52 seconds! It moved faster and achieved a slightly better result in regard to artifacts.

Denoise module in RX 3 advanced stand alone

Another major enhancement in RX 3 is that the Denoise tool has been updated: Spectral and Dialogue tabs cater to your applications. The new interface upgrades help make choices easier when you're fixing dialog or vocal tracks that have background-noise issues, or if you need to step into a deeper set of options for more complex situations. This arrangement displays auto and manual features for dialog as well as an advanced-settings tab for spectral. Please note that these changes are not just skin deep, they are two entirely different algorithms catering to the different applications. When opened as a plug-in in your DAW, they will show as two different plug-in choices.

These design changes open the door to video editors with less of an audio background, but still offer depth of control for audio professionals, all in one simple tool, laid out in straightforward fashion.

In processing a few examples, I found the engine only slightly faster, as my clips were short. However, the results possessed even fewer artifacts than in the previous version of RX. For my sample test of dialog in a noisy environment, I used audio recorded by our B&H video team, on two separate locations. First, we used the Zoom H6 with the MSH-6 mid-side microphone attachment, inside the B&H SuperStore in NYC, adjacent to one of our conveyor systems. In this example, I ran the Denoise module through the dialogue setting and engaged the auto mode to see how well it could clean up the audio on its own. I only Denoised the first track (Mid), since I would discard the second track (Side) when mixing. With only minor tailoring of the threshold and reduction settings, the results were decent. You can hear the suppression of some of the more cycling background noise as opposed to the more random, intermittent sounds. This is a very effective and fast process with the self-detection tool.

Zoom H6 Denoise example

For the second dialog Denoise example, I took a sample from the same team using the Shure VP83 Lenshopper on the streets of Manhattan, adjacent to a park and a street bustling with traffic. In this situation I still used the dialog setting. Instead of engaging auto mode, I directed the module to learn the noise, and I crafted my settings to find the balance I was seeking. The results, again, were great, and this time I'm not only hearing less clouding of the the dialog by background noise, I'm also getting more of the birds from the park, thereby engaging the audio to truly complement the video. In this example, you will hear me increase the reduction level aggressively during processing, so even more noise is suppressed. You may also notice the small level of dialog artifacts from the processing coming out. One powerful difference to be aware of in dialog mode as well is that it is real-time and zero-latency, so when editing to video, this tool will not cause audio placement issues.

Shure VP83 Denoise example

I'd like to point out that with RX 3 Advanced, or any other noise-reduction software, there is no magic tool on the market which will remove all noise, yet keep all your target audio completely intact at full fidelity. You, as a user, have to sacrifice some degree of fidelity in your final audio to gain the desired result, and you may have to process audio further after repairing it.

Next, I wanted to test the updated Declip tool to repair audio, which had been intentionally overloaded with a mic preamp. For many repairs, it does what you thought impossible—it recreates the digital audio that was leveled to a square wave at peak, and estimates what was there. This impressive module has also had a facelift since the RX 2, and now uses nice, large sliders that create a more "audio-equipment feel," rather than the previous version's bright blue ones. There is also a visual representation of the wave in the Threshold setting, as well as the new "Suggest" setting that saves you time by letting the software decide which settings are ideal for correcting the issue.



In my example, I took a simple voice-over test, recorded with an AKG 414 through a Focusrite Forte interface, connected with a Kopul 4000 cable. I ran the input hot, intentionally, to peak the audio-out, resulting in a distorted waveform in my DAW. In the before and after screenshots (which can be exported right out of RX 3's File menu, by the way), you can see the recreation of the damaged wave being redrawn. Now, when you listen to the sample, you can hear the damage, and the repair—with no artifacts from the processing. You can easily conclude that this tool can be a lifesaver, as it can correct those one-time-only recordings and make them mix-worthy.

Moving to the next test with the same mic and preamp, I recorded a snare drum and dialog from roughly 10 feet away, thus creating a room sound, to try the new Deverb module. The room is almost all drywall and tile, so this would be a reliable demonstration of bright room ambiance. Removing reverb is very difficult, and few tools do it correctly. It is a common issue to have a dialog or vocal track recorded in an untreated room with poor microphone placement, leading to a track that's almost unusable, and I wanted to see how the RX 3 Advanced would fare against such a challenge.

Dereverb module example

You can see from the process window, in the first group of waves, the additional dereverb noise. Then you can see it being reduced significantly in the second group. While the audio does have a few phase-adjusted artifacts, you can hear a much tighter recording with less reverb.

For my last example, I wanted to try out the new Extract Center tool in the Channel Operations module. This is a very special tool, in that it offers a mathematical  process that can collapse just the sides of your mix, or the center, at your set degrees. It can work wonders for several issues: pulling a vocal a little further out from a song, post-mastering, narrowing a mid-side mic back to the center after mixdown, and even extracting a vocal from a track for remixes. Taking a clip from Third Kind's "Hurricane," I set the threshold settings to be a little aggressive, to see how much of the vocal I could get to stand by itself. The results were impressive. As you can see from the result below, and hear in the sample, the audio has sacrificed very little fidelity, yet the guitars are considerably attenuated.

Channel Operations

Extract Center Example

While all these tools may at first appear only useful when problems arise, these few demonstrations display that anyone engineering in audio for video, post production, or music recording can benefit when that curveball gets thrown your way. RX 3 Advanced has capabilities that stretch much further than we have demonstrated in this hands-on review, but the major highlights show some great new features that make this product stand out.

Repair tools may not sound as exciting as the next hot virtual analog synth plug-in, but it is the tool that will save you that extra take when you need it to, and most likely, save a lot of money. You can be assured that in professional applications, this software can pay for itself in the first take or two where you need to "make it happen."


For more information, stop by the B&H SuperStore in New York, speak with a sales professional on the telephone at 1-800-606-6969 or contact us online via Live Chat.