iZotope has released the latest version of their iconic audio restoration software RX 9. In this article, we have all the details of what is new, an exclusive demo video, audio examples and we tell you what we think of RX 9 from iZotope.

What iZotope Say About RX 9 And RX 9 Advanced

iZotope Principal Product Manager, Mike Rozett told us…

“For RX 9, we have revisited some of our most important processing tools and made them even better. We wanted to focus on the fact that dialogue is getting noisier and noisier: from location shoots to warehouse sets, to wireless interference on sound stages, to ADR that’s being recorded remotely in cars and closets and kitchens instead of in studios. The industry is facing more and more noise, with less time to fix it. We’re here to help.”

With that underlying objective with iZotope’s intentions for RX 9 Let’s look in a little more detail what new features iZotope has brought to RX 9 and RX 9 Advanced.

• Enhanced separation with Dialogue Isolate [Advanced only]: Using iZotope machine learning, this new version of Dialogue Isolate makes it easier than ever to extract clean dialogue from its environment, without artifacts, enabling you to be able to instantly separate speech from crowds, traffic, footsteps, weather, and more. You can also still use the old algorithm by selecting the ‘Use Legacy Algorithm’ option. Dialog Isolate works in the Standalone RX 9 Advanced app and as an AudioSuite plugin.

• Complex Ambience Match [Advanced Only]: Ambience Match now has a Complex Mode that allows you to capture real background movement and texture from production audio. Create complex fill from non-stationary ambience with control over what and how much you include in the fill. You can match scenes featuring rain, traffic, crowds, and more! This works in the Standalone RX 9 Advanced app and as an AudioSuite plugin.

• Dynamic De-hum [Standard and Advanced]: Use the new Dynamic Mode in RX 9’s De-Hum to remove any amount of hum, wireless ring, or interference, without artifacts or affecting the audio quality. In just one pass, you can now eliminate radio and transmitter interference, camera and generator hum, and more. This works in the Standalone RX 9 Advanced app and as an AudioSuite plugin.

• Restore Selection[Standard and Advanced]: A new function of the History Pane enables users to restore only a selected region of audio to a previous state. For example, imagine that you've run Mouth De-click on a long piece of dialogue, only to find that there are some places where you wish you hadn't removed the mouth clicks. This feature allows you to make a selection, then choose a step in your history to "go back to" for only that selection.

  When hovering your cursor over an item in the History Pane with a selection made in the spectrogram, an icon appears in line with the text. Clicking this icon will restore your selection to that point in history.

• Expandable History list[Standard and Advanced]: The History Pane is now expandable, so you can see more of your edit history without scrolling. This enables users to more easily see up to 30 steps of undo history for quick reference.

• ARA Spectral Editor [Standard and Advanced] Logic Pro X Only: Clean up unwanted sounds directly in Logic Pro X using the new Spectral Editor, which puts the power of Spectral Repair inside Apple’s Logic Pro X using Audio Random Access, better known as ‘ARA’. This means that Logic users don’t have to leave the DAW to use Spectral Repair to remove and repair unwanted noises in dialogue and instrumental tracks. You will be able to see your audio via the Spectrogram in Logic with access to the Attenuate and Replace modes and includes expandable & collapsible advanced controls and all of the selection tools available in the standalone RX application. In fact this feature is already available in the subscription version, RXPro.

Learn More

You can learn more about these new features and watch demo video from iZotope on the iZotope RX 9 Product page or scroll down to see how we got on with the new features in iZotope RX 9.

What We Think of iZotope RX 9 And RX 9 Advanced

In this video, Mike demonstrates his experiences with the 3 main new features in iZotope RX 9, the new version of Dialog Isolate, the new version of Ambience Match and the new version of De-hum.

Both Mike Thornton and Tom Lowe have been trying out iZotope RX 9 and this is what each of them have found, without comparing notes, giving you two perspectives on RX 9…

Mike Thornton

I have been working with iZotope’s audio restoration software since RX2. Each time they release a new version, things that were difficult or impossible to do become possible or much easier. Let’s find out what there is in RX9…

Dialog Isolate

Let’s start with Dialog Isolate. At first glance, it may not look any different, but on closer inspection, the bottom drop-down is different, instead of choosing different separation algorithms, it now selects the quality level with a choice of Good and Best.

All the other changes to Dialog Isolate are under the hood. It has a brand new machine learning neural network, designed to separate dialog more effectively than the previous version used in RX6 to RX8.

So let’s try it out, here is one of my classic test files recorded at NAMM in 2012…

Note that I am going to deliberately push this harder than I might on a real job, to see what it can do. As before there is a Sensitivity control that allows you to set what is dialog and what is noise.

Once I was happy with the settings, the first thing I noticed is that the new algorithm is significantly slower than the older one, typically at least twice as slow. If you do find it too slow then iZotope has made the older algorithm available. By clicking on the gearwheel there is an option to select ‘Use legacy algorithm’.

Here are two versions, one using the old algorithm and the other the new algorithm…

You should be able to hear that although the old one is pretty good, the new one is better, it gets rid of the last bits of the background.

Whilst we have Dialog Isolate open, I want you to listen to this example…

As you can hear a text alert went off during the recording. Let’s compare the old and new Dialog Isolate algorithms and see how they handle this phone interruption. First the old algorithm…

It doesn’t really touch the text alert sound. Now the new Dialog Isolate algorithm…

The phone text alert hasn’t gone completely but it’s pretty close and so much easier to achieve than what I used to have to do.

Job done.

Ambience Match

Next, we are going to take a look at the new Complex Ambience Match option, which is designed to capture real background movement and texture from production audio and create complex fill from non-stationary ambience with control over what and how much you include in the fill.

The original ambience extract option is great but the criticism is that the audio it produces is a little flat, and static. So iZotope set about improving it so that it extracted ambiences could have some variety in them. So let’s take a look. Here is a clip with some great ambience of a New York street…

In the RX 9 Ambience Match module, we now have two options, Static, which is the original algorithm and Complex which is the new one. Let’s just remind ourselves of the old one…

There you have it, yes it’s a bit static but it enables us to create clean fill where there wasn’t any.

When we switch over to the Complex option are 3 sliders; Threshold, Movement and Randomness, which are made active.

iZotope recommends that for the Dynamic mode you do not select the whole file but select segments to provide the basis or a new complex fill. Let’s see what we have got by doing that…

It has variation but there is discernable looping. Before we do anymore, let’s look at what the two controls down the bottom do. Movement looks for transient or random events and removes them from the selection. Whereas with the Randomness control, negative values will loop as much as possible the original selection and positive values will randomly shuffle what has been captured.

But what if you haven’t got enough clear ambience to use? There is one more option with the Complex Mode and it is intended to be used when there isn’t enough clean ambience to work with. iZotope has added a Threshold control, that adds a ‘strip-silence’ option so that you can use the Complex mode on complete files with dialog. You will need to adjust the Threshold down as low as you can get. Note that if you go too low it will report that it is unable to create an ambience file. Here is an example from the same file which had speech included in the selection, and then the Threshold adjusted so that the algorithm ignores the speech.

It still isn’t perfect, I can still hear looping. This is not the only example I have tried the Complex Mode on, I have tried a wide range of material and I am yet to find something that works. So for me, it’s back to the Static mode for now. It’s a shame that iZotope aren’t able to ignore the speech more effectively. After all, with the new Dialog Isolate algorithm, shouldn’t it be possible to identify the speech and remove it and use that to create a complex fill file?

De-hum Dynamic Mode

Let’s move on to the new Dynamic De-hum. You might remember this example…

The original De-hum module, which was getting a little long in the tooth, other than a doubling of the number of harmonics was OK for static hum, preferably when it wasn’t in the same space as the speech that you want to preserve. Also when working with a file with 50Hz or 60 Hz hum, even with the 16 static notch filters, don’t reach 1kHz. With files like this one, the buzz elements can go way beyond 1kHz as you can hear in this example using the old static De-um option…

As you can hear it’s taking care of the low-end hum but not all the buzz higher up. This is where the new Dynamic mode comes in.

Instead of 16 notch filters, with RX 9 there are 1024 and all those 1024 are dynamic notch filters, which can change the amount of reduction adaptively, based on the level of the incoming signal at this particular frequency. So let’s see how this handles this example. With the new mode, you do need to use the Learn option and select a clean section of the buzz without any speech…

It allocated all those extra notches to each one of the frequency components in the buzz and because they are dynamic filters they don’t eat into the wanted audio as the static filters do and there is no buzz left.

Job done.

Because the new dynamic mode can hand frequencies all the way up to 20 kHz I wondered where this module could have other uses. Have you ever had to deal with a recording with a PA system on the edge of feeding back like this one…

By highlighting the section where the PA is rings and starting to feedback we can teach the dynamic mode to add notches to cover the ringing and feedback…

However although it has made a good attempt at fixing it, it isn’t perfect. To fix it I adjusted the Q of the dynamic notch filters from the default Q of 1000 down to just under 260 and reprocessed it…

There you have it. Job done.

ARA Spectral Editor For Logic Pro X

It turns out that this feature is already available in the subscription version RX Pro. Spectral Editor is only available in Logic Pro as an ARA 2.0 plug-in. You can add RX Spectral Editor (ARA) to your project as long as you are using the latest version of Logic Pro. Note that the RX Spectral Editor ARA plug-in is supported in Logic Pro 10.5.1 or higher.

You then need to add RX Spectral Editor (ARA) plugin on the first insert on an Audio track. Next, select a clip in Logic and start playback. The selected audio will appear in the spectrogram view of the Spectral Editor (ARA) plug-in after starting playback. Make a selection in the spectrogram view of RX Spectral Editor of the problem audio you want to remove.

You can then process your selection, but the workflow is slightly different depending on whether you have the Advanced Control panel collapsed or expanded…

• If the Advanced Controls Panel is collapsed: click the Attenuate or Replace button to apply the associated processing to your selection.

• If the Advanced Controls Panel is expanded: click the Process button to process the selection using the currently selected Mode.

To work on another clip in your session, select it and start playback. Once playback has started, you will see your new clip load in the plugin’s spectrogram. Trimmed clips appear the same way in the spectrogram as they do in Logic.

There are two caveats using the ARA Spectral Editor in Logic Pro X…

1. Attenuate: Horizontal, Attenuate: 2D, and Replace mode do not support processing selections longer than 10 seconds. If your selection is longer than 10 seconds, the processing will automatically default to Attenuate: Vertical.

2. ARA plug-ins are updated when playback is started. If you trim or otherwise change a clip that was already loaded in Spectral Editor, the spectrogram will not be updated until you’ve stopped and restarted playback with the clip selected.

It’s nice to see iZotope support ARA, but why just Logic Pro? Why not other DAWs that support ARA like Studio One from PreSonus or Nuendo from Steinberg?

Conclusion

There is no doubt that RX 9 is a ‘make-it-better’ update, rather than adding a raft of new features and is squarely pitched at the audio post-production sector and making our lives easier. For me, it feels like RX 8.5 rather than RX 9. That said, the dynamic De-hum is the star of the show, followed by the selective undo feature where you can choose to undo processing within a selection. This is great if you have processed a long file and then find that there are a couple of sections where you have either overdone or underdone the processing, you can go in, undo the processing just for that selection, and then take another swing at it.

The Dialog Isolate improvements are very helpful too, but it’s a shame about the Ambience Match Complex Mode, which I have not found to be successful with anything I have tried it on. It’s a great idea but in my opinion it still needs work. So let’s hope iZotope can improve it and release a free incremental update.

Tom Lowe

Having been a user of every version of iZotope RX since version 3, I've processed thousands of hours of audio through it and have become very adept at knowing what tools suit specific tasks (it's not always the tool that's designed for that job either).

Dialog Isolate

One of the additions in recent years was Dialogue Isolate, and whilst I'd had some success using it, it wasn't often my first choice. I was aware of what it could do, but in my experience, it was very easy to over-cook and getting audible gating artifacts in the resulting audio. As a result, I stuck to my previous methods in RX and introduced other restoration tools into my workflow. However, there's a high chance RX 9 has completely changed my opinion on Dialogue Isolate.

Recently, I was tasked with cleaning a piece of audio from an archival 16mm print, where floodwater had damaged the film stock. The damage was so bad that the emulsion had started to come away, affecting both the picture on the film as well as the optical soundtrack. The result was a crackly thump throughout the 5-minute film. Unfortunately, this project is under an NDA, so audio examples aren't possible; however, screenshots of the spectrogram before and after give some idea of what I was facing.

The spectrogram shows a range of issues with the audio, but the "thump" itself isn't particularly clear to see. The other problem was that it occurs during the dialogue and is of an equally audible level. Because of this, existing tools such as Attenuate and Spectral Denoise weren't going to be particularly helpful.

My thoughts immediately turned to the newly-improved Dialogue Isolate module. My way of using RX is always to do multiple passes, each removing a small amount of noise rather than one big pass. I set the Sensitivity slider to around 4 or 5 and the noise attenuation at -4dB. After processing, I listened to the result, and although there was an improvement of perhaps 10%, I knew I needed to do more. So, I did a second pass using exactly the same settings, expecting another subtle improvement. Upon playing back the audio, I had a jaw-dropping moment. The audio I was listening to was now at least 70-80% better than the unprocessed version.

I could have gone back and tried to do the same in one pass or done an additional pass using the same settings, but I knew after this much improvement, anything else risked ruining it. I had the perfect result where the audio is clear and very audible, with no apparent artefacts. The background hiss and crackles had also been removed, meaning I didn't have to do any additional processing to remove these.

I was fortunate that the audio was predominantly dialogue, and as the name implies, this module is designed precisely for this type of processing. Had this been a piece of music or other audio, I don't think I'd have been able to get the same results. What impressed me the most is it worked with band-limited, 50-year-old (or thereabouts) audio from a badly damaged 16mm print. The result is that a piece of film considered ‘lost’ is now fit for broadcast and public consumption.

Here is another example with audio I can share, showing the difference between the two quality settings; Good and Best. To start with, here is the original unprocessed audio for reference…

Having determined the most appropriate settings, here is the processed version on the Good quality setting…

and to complete the set, here is the same audio processed on the Best setting…

Ambience Match

Ambience Match has a new feature where it can create variations in the ambience it creates. I used this a little bit, and I feel it needs more work. One of the nice things about this module, from when it was first introduced, is that you can give it a piece of dialogue and it will ignore that and extract a continuous piece of clean ambience, without any dialogue. The problem is that the result can sound very static.

This updated module aims to overcome this and create a more natural piece of ambience with variations. The problem is, when you feed it a piece of dialogue, the algorithm will now include parts of that dialogue in the resulting ambience. You now have to give the updated algorithm a piece of audio without any dialogue from which to learn.

This module exists because when editing dialogue, you often don't have much clean room tone or ambience in the first place. I think you can see the problem here!

Dynamic De-hum

The other update is the Dynamic option added to the De-hum module, and this is the first update to this module since RX was developed. The competition started to catch up with RX on this front, so this is a welcome update.

ARA Spectral Editor

One completely new feature is ARA integration with Logic for the spectral editing tools. As I don't use Logic, I was unable to test this, and it feels a slightly odd inclusion as I'm unaware of many people using Logic for this type of work. It feels like a toe-dipping exercise preparing for ARA2 integration with other DAWs in RX 10.

Conclusion

Overall, I'd say RX 9 is an evolution and not a revolution, but that's no bad thing. Making existing tools better and making new things possible is no bad thing, and I'd rather have stable software with useable features than a ton of new features that don't work particularly well or that I don't need in my type of work.

iZotope RX 9 System Requirements

Operating System

• Mac: macOS Mojave (10.14.6) - macOS Big Sur (11.6)*

  • Intel Macs or Apple Silicon Macs in Rosetta mode only.

• Windows: Windows 10

Plug-in Formats

• AU, AAX, VST2, VST3.

• All plug-in formats are 64-bit only

Supported Hosts

• Apple Logic Pro

• Avid Pro Tools 2021

• Ableton Live 11

• Steinberg Cubase 10.5-11

• Steinberg Nuendo 11

• PreSonus Studio One 5

• Cockos Reaper 6

• Fruity Loops Studio 20

• Adobe Audition CC

• Adobe Premiere Pro CC

RX 9 Pricing & Availability

RX 9 is now available from iZotope and from selected retailers

• RX 9 Standard: Regular price $399, Introductory price $299

• RX 9 Advanced: Regular price $1,199, Introductory price $799

• RX Post Production Suite 6: Regular price $1,999, Introductory price $999

The iZotope Post Production Suite 6 includes RX 9 Advanced, Dialogue Match, Insight 2 metering, Symphony 3D and Stratus 3D reverbs by Exponential Audio.

Upgrades

Existing iZotope customers are eligible for additional loyalty discounts on RX 9 and RX Post Production Suite 6. Details can be found in the user’s account section on iZotope.com.