iZotope RX11 Advanced

RX11 brings a number of improvements to both the Dialogue Isolate and Dialogue Contour modules for processing voice‑based audio.

We check out the latest iteration of what arguably remains the most comprehensive audio repair software available.

Thankfully, spectral editing is not what it used to be! Yes, you can still do those forensic‑level manual edits within the spectral display where that’s what’s required and if you have the skills, but with easy‑to‑use processing tools and some clever machine‑learning algorithms under the hood, products such as iZotope’s RX now allow anyone to access the power of this kind of editing.

RX remains one of the leading products in this field and, some 18 months after Hugh Robjohns assessed RX10 (SOS January 2023), iZotope have now delivered RX11. As before, RX11 is available in three versions, Elements, Standard and Advanced, that at heart are the same app but with more or fewer features for different prices. For full details on those differences, you can check out the comparison chart on iZotope’s website, but the main focus of this review will be the ‘full fat’ Advanced version.

iZotope have presented recent iterations of RX as a ‘toolkit for audio repair’, and that toolkit comprises a collection of processing modules, each designed to streamline a specific editing task, such as noise reduction, de‑clicking, reverb removal, reducing guitar finger squeaks, matching EQ from a source to a target, or stem unmixing/rebalancing. The quality of result can, of course, depend on the nature of the audio being processed, but more often than not RX is capable of dramatically improving poor audio, and even nudging ‘unusable’ audio over the threshold into usable territory. As well as building on the functions of RX10, RX11 integrates further advances in machine‑learning technology and adds some headline ‘new and improved’ features.

Voice Control

RX11 brings a number of new and improved tools for processing audio containing the human voice, and perhaps the most significant is a reworking of the Dialogue Isolate module. With machine learning underlying the processing, this now combines both de‑noise and de‑reverb options into a single, simple‑to‑use module that can be applied in real time (the modest amount of latency is perfectly acceptable for post‑production editing). What’s more, this facility is, for the first time, available in RX Standard (using the real‑time algorithm), while RX Advanced’s version adds a further high‑quality (albeit offline rather than real‑time) option as well customisable multiband processing.

In my tests, Dialogue Isolate did a remarkable job of clearing away the sonic clutter in some old‑school dialogue recordings and, importantly, it did so without compromising the quality of the spoken voice. There are obvious applications in journalism and broadcasting, but I can also easily imagine podcasters using Dialogue Isolate as part of their post‑production workflow for that final bit of clean‑up. That said, if this is all you want a tool to do, there’s plenty of competition, including iZotope’s own VEA and Accentize’s dxRevive (with the latter being particularly impressive).

Available in RX Advanced’s standalone Editor (not as a separate plug‑in), the Dialogue Contour module has also been improved. As well as a reorganisation of the various slider controls, this includes a pitch curve that’s superimposed on the waveform, giving you a very useful visual representation of the rise and fall of ‘expression’ within the dialogue, and as you add/edit pitch‑curve nodes in the spectral display, a second curve shows how your changes affect the pitch curve. The control set also now includes a Variation slider that allows you to either expand (more expressive) or contract (less expressive) the overall pitch variation in this curve. This can be really helpful if you need to finesse the expressivity in dialogue or, for example, smoothly drop in word edits from multiple takes or different points in a recording.

The Repair Assistant also benefits from improvements. This is available in all editions of RX but, of course, the signal chains it creates for you from the initial ‘learn’ process is restricted to the module components of the version you own. You can choose between processing voice (dialogue or sung) or instrument sources and, as shown in the screenshot, the module/plug‑in then provides a compact set of initial controls across four panels. If required, you can open the full module chain for more detailed control. While this approach is very flexible, it does seem fairly CPU intensive. So if working on a modestly specified machine, you’ll probably want to render your Repair Assistant edits before moving on to the next editing task.

The Repair Assistant has also been improved in RX11.

RX10’s global Separation slider has been replaced by individual Sensitivity sliders for each of the four stems (Vocal, Bass, Drums, and Other).

Unmixing Magic

The new Sensitivity controls allow you to encourage Music Rebalance’s sophisticated stem unmixing algorithms to focus on a specific component within the separation process.Music Rebalance, RX’s stem separation tool, also benefits from improved machine learning and there are some useful refinements to the GUI and control set too. RX10’s global Separation slider has been replaced by individual Sensitivity sliders for each of the four stems (Vocal, Bass, Drums, and Other), so you can now fine‑tune the relative emphasis the separation algorithms place on these four sounds. In practice, this translates to the user enjoying an extra degree of control that means you can focus on a specific target should you wish.

Creating instrumental mixes or backing tracks (in other words, removing vocals) is almost always achievable to a remarkably high standard. More problematic – even if it’s still something approaching sorcery! – is the opposite: isolating a vocal to create an a cappella. Busy mixing, lots of reverb/delay on the vocal, or the presence of backing/harmony vocals, make generating a clean result more challenging. However, the Sensitivity slider does give you some room to experiment on that front and, given the additional voice repair options RX offers, there are options for some further cleaning (de‑reverb can be really helpful, for example). I should mention that isolating harmony and backing vocals remains a challenge: whoever cracks that particular bit of machine learning/AI first will undoubtedly jump ahead of the competition. And speaking of competition, while RX is impressive it’s not the only software tool capable of this sort of unmixing magic; I’d say that the likes of Steinberg’s SpectraLayers and Hit’n’Mix’s RipX are capable of comparable results.

Central Perk

There’s one further new feature in RX11 that might also have some potential when it comes to vocal isolation, among other things: the addition in both the Standard and Advanced versions of a Mid‑Sides editing and processing mode. For stereo files, in the View menu you can now flip the channel mode between the conventional Left‑Right stereo and Mid/Sides. Once you’ve done that, all RX’s modules can process the Mid and Sides components of a stereo signal separately. This could open up all sorts of applications, but when it comes to isolating a vocal you’ll often find that the lead vocal sits mainly in the Mid channel, while the Sides contains a little more of the ambience, effects and backing vocals. It might be a marginal gain but if you’re looking to extract a vocal part for a remix, this would be well worth exploring alongside Music Rebalance.

There might also be potential here for getting creative with the bass or drum stems created by Music Rebalance, whether it’s to refocus bass frequencies in the middle of the stereo image, or to add a little extra stereo sparkle to a drum stem. The same Mid versus Sides content variation might also apply with dialogue in stereo field recordings (particular noise elements might well be more prominent in the Sides). Anyway, it’s a great addition and will, I suspect, get lots of RX users thinking of new ways it might help with routine repair and editing tasks using the existing suite of modules.

Streaming & Loudness

RX10 Standard and Advanced included a Loudness Control module, while Advanced also offered the Leveler. These modules remain and each bring something to the table in terms of loudness management but in RX11 Standard and Advanced iZotope have upped the game considerably with Loudness Optimise and Streaming Preview.

The new Loudness Optimize module lets you easily apply upwards compression to the quieter sections within your audio project to help avoid any undesirable adjustments being made by subsequent loudness normalisation.

Loudness Optimise provides a visual and numerical summary of a file’s loudness. If required, it can then apply upward compression to the quieter sections to increasing the average loudness while leaving the louder sections and peaks unaltered. The display includes a representation of the ‘weighted gate’ that’s used in the BS.1770 Integrated Loudness calculations. Any audio that falls beneath this gate is automatically excluded from the integrated loudness calculation (which is why, for some audio files, the ‘% measured’ value shown top‑left will be less than 100%). In calculating average loudness, the gate means sections of silence or low‑level material, not least within TV/film broadcast content, do not cause ‘misleading’ loudness measurements, ensuring a more consistent result for the listener.

For music tracks aimed at streaming platforms, though, where significant passages in your track fall below this gate, this can have a marked impact on how the loudness normalisation of each streaming platform is applied to your track. Loudness Optimise therefore provides a great way to spot this potential problem. Then, through a combination of its various controls, you can apply some upward compression and then preview (and render) a modified version of the track that’s less likely to fall foul of the loudness normalisation applied during streaming. It’s really easy to see and hear what’s going on.

Streaming Preview complements this very nicely. It allows you to audition your mixed and mastered track with the loudness normalisation processing that would be applied to it by specific streaming platforms. You can also render a version of it that includes that processing. Currently available presets include variants of Amazon Music, Apple Music, Spotify, Tidal and YouTube, but you can configure a specific loudness, codec and audio quality combination manually if you need to. It’s a super‑useful tool that allows you to determine whether, in terms of its loudness, your track is ready for upload or still requires some finessing.

The new Streaming Preview module lets you hear how your music will sound once subjected to loudness normalisation on any of the major streaming platforms.

Take It Up To 11?

So, is RX11 for you? Well, as Hugh Robjohns suggested of RX10, it’s fair to describe this as a case of evolution rather than revolution, but it’s worth reminding ourselves just how remarkable RX11 has become. Yes, it’s a specialist tool that not every audio editor or music creator will feel compelled to use, and yes, there are some apps that might outperform it on some individual tasks. But if your work routinely involves rescuing audio that is sub‑optimal, RX is one of the very few comprehensive options that you might turn to in order to save the day. In RX11, iZotope have pushed the limits of what’s possible that bit further and made doing it that bit easier. For many pro audio editors regularly trying to do the impossible to a tight deadline, that may be all that matters.