By Cory Choy

Full disclosure: I am a big iZotope RX fan. Why? Because out of all the tools that have been made accessible to the post world over the years, only a few have been complete game changers. iZotope RX is one of those tools.

Cory Choy

The elegance and ease with which one can view and edit the spectral representation of a waveform is pretty much unmatched by any other software, and the importance and power of that cannot be emphasized enough. Not only that, but iZotope’s suite is available at a very reasonable price point (with incredibly affordable sales offered throughout the year) and it does not require the dreaded iLok to use.

There is also something to be said about the many different ways iZotope allows one to go about restoring audio, from Spectral Repair to Dialogue Isolate to Voice Denoise to Spectral Denoise — and each one of these tools goes about things a different way and helps tackle different problems.

So when I was given an opportunity to put the new version of iZotope RX through its paces, I jumped at the opportunity — knowing that even though I was a fan of the company, that my review would be thorough and honest.

For this review, I will be focusing on the suite’s new features from the perspective of a sound designer/mixer. At the end, I will give my thoughts about whether it’s worth upgrading from an earlier version and whether it’s worth buying for the first time.

Repair Assistant
The new version of iZotope RX includes a new and improved Repair Assistant. This is an automated tool that analyzes a piece of audio and gives recommendations on how to make the audio sound its best — mainly by subtracting unsavory elements. As you can see in the image provided, it uses De-Ess, Voice Denoise, De-click, De-Click and De-Reverb. That’s a lot of De’s.

I am always skeptical of tools like this, as I think it ends up being faster for someone like me who knows after listening what kind of tools need to be used. However, if you don’t know what these various tools do, and you don’t have experience using them, I could see it being nice to see what certain settings do. The audio clip that I provided was a remotely recorded clip for a podcast. The high-end was very much lacking and the voice was thin and compressed sounding.

I selected two minutes of audio and hit “learn” on Repair Assistant. It thought for a long time (almost 1 minute and 30 seconds) and after listening to the result, I wasn’t particularly impressed. All it really did was subtract the smallest amount of ambient noise. The voice still sounded… not good. I guess this brings me to what I think would make repair assistant better — I wish it used all of the modules rather than just a few of them. With my prior knowledge of RX, I know that the new and improved Spectral Recovery would be the thing that would make the most difference here, so I gave that a whirl… And that was also a pretty time-intensive process.

Spectral Recovery
What Spectral Recovery does is try to interpolate what’s there and add back the high-end that was subtracted from the VOIP. Pretty cool stuff.

Here is the new waveform.

WOW. What a difference! It sounded so much fuller! When I A/B’d it with the Spectral Recovery from RX 9 Advanced, I would say it was noticeably improved. Unfortunately, however, it was unable to add back enough information from the 6KHz-8KHz range, so it gave the speaker a lisp. I decided to try Repair Assistant on that and still didn’t get that great results. Do you know what would work here? A Re-Ess function. I wonder if that exists…

With a lot of fussing around, I think I could probably improve the sound from here about 30%-50%. But it would be nice to have a preset just nail it for me. Still, the new Spectral Recovery was pretty impressive.

Text Navigation
What does Text Navigation purport to do? It allows you to “see what you hear with the new Text Navigation function. It analyzes dialogue and displays a text transcription above the spectrogram that’s in sync with the corresponding audio.” Whoa! IZotope has officially entered the automated audio to transcription realm [putting it in direct competition with the likes of Otter and Trint and Descript. That’s pretty cool… if it works. Let’s find out.

The first question I have is, where is the darn thing? I looked on the right-hand side toolbar and I can’t seem to find it. Let me check the menus…

Not there… how about Help? Nope. Nothing useful there either. Gonna have to Google it. There’s a video I think… Aha! It’s here only on the bottom left:

I had a hard time finding that. So now I clicked the icons and… lanes are there but empty

How do I populate this? I guess more YouTube for me… (I really think this should be in the manual.) Oh! Turns out that it is transcribing the entire file, and not just the selected audio. That will take forever, as this clip is three hours long. Let me make a subclip and get back to you.

Okay, lots of thoughts here. This feature was hard to find, and once I found it, it wasn’t super intuitive. There wasn’t enough in the manual. The transcription seems to take a while and right now it doesn’t seem super-accurate on a mediocre recording. There is only one speaker, but manual speaker detection seems to think there are two. I can’t get an easy way to look at the transcription… shouldn’t it populate on the left? Is there a way to export the entire transcription as a text file?

The sentence I hear with my ear is, “Great, thanks Yehuda, happy to be here.” The only words that transcription got right were “be here.” Let’s try with a better sounding sample file.

Hmm… it’s better, but with only about 60% accuracy. Not good enough. However, the search function is really cool! I search for the word “all” and it allows me to instantly navigate to those parts of the timeline. One problem? It only found 2 of 4 instances of the world “all” in a 60-second spot.

Conclusion? The intent of this module is really, really cool, but it needs to be refined. If the transcription isn’t more accurate, the search function won’t be as useful as it could be. The way the transcript is displayed is inelegant and clunky. iZotope absolutely needs create a simple and easy way to export the transcript as a text file. I’m excited to see where this goes, but with the experience I just had, it feels more like a time-suck than a time-saver at the moment.

This goes for the Multiple Speaker Detection as well. Not accurate enough to be super useful for me… at the moment.

Okay! What’s left…

Dynamic Adaptive Hum Removal
This was good. It was also fast, not super-slow like some of the other functions.

Before

After

Result? Multiple bands of hums greatly reduced and when it switched frequency in the clip, the DeHummer kept up. Good and fast! Dialogue still intact. Impressive! To be fair, for this instance, I could use RX9 and press the button twice instead of once and I think it would sound just as good, but I could see for a narrow set of applications that this would be real time saver.

The Verdict
Is it worth upgrading? If you already have RX9 Advanced, in my opinion, no. RX9 can do 95% of what this can do well. If the auto-transcription function was better and more accurate and allowed you to export the transcription as a text file, I would say it would absolutely be worth upgrading. Maybe if there is a patch?

Is it worth buying RX for the first time? Yes. Absolutely! These tools are incredibly powerful and affordable.

I’m so excited to see what comes next from these folks.

Cory Choy is an Emmy Award-winning sound mixer, producer and director of Esme, My Love. He is the owner of NYC- and LA-based Silver Sound and proud father of Grace Lee and Jakob Fox.