RX 11 Dialogue Isolate: Lives Up to the Hype?

Alright folks, buckle up because I'm diving deep into iZotope RX 11's new Dialogue Isolate plugin. We're talking real-time noise reduction specifically for unscripted content, my bread and butter! 

I couldn't hold in all my thoughts as I worked on a competition cooking show and got to take RX 11 Dialogue Isolate for a spin! I hate reviews that show its noise removal job; I want to know how exactly it can aid my mixing workflow.

Let's dive in!

A Familiar Face with New Tricks

First impressions? The layout feels instantly comfortable.  

It mirrors the structure of other noise reduction plugins I already know well, like Waves Clarity VX Pro and Accentize's dxRevivePro. 

However, at first glance, it looks most similar to the recent hot commodity noise reduction plugin on the market: Supertone Clear. 

When you first load it up, it has 3 faders: one for dialogue, reverb, and noise. For us seasoned editors, that translates to less time fumbling with knobs and more time making magic happen. 

There's a sensitivity level, which aims to determine how surgical the separation between noise/reverb and dialogue.

At the bottom, you'll find two Quality drop-down options: Good (Real-Time) and Best (Offline). These are not any different from the initial sentiments.

But, now there's a little arrow next to it, that when clicked, opens up a spectral view with 4 adjustable frequency bands to control precise noise reduction, unlike Supertone Clear, but like Clarity VX Pro and dxRevivePro.

Izotope really said you get the best of both worlds! I applaud the simplicity while maintaining the specificity that we pro mixers need. 

Natural-Sounding Noise Reduction: A Breath of Fresh Air

Let's talk about the real star of the show here: the noise reduction itself.  

Unlike some plugins that leave your dialogue sounding like it went through a meat grinder, the new Dialogue Isolate algorithm delivers a refreshingly natural sound, even with light settings applied.  

This is absolutely crucial for unscripted content, where preserving the authenticity and nuance of the dialogue is paramount.  

No one wants to watch an interview where the actors sound like they're auditioning for a robot uprising!

Here's a specific example that had me cheering: RF interference.  

That high-pitched hiss that loves to haunt wireless lavalier recordings?  

Dialogue Isolate cuts through it like a hot knife through butter, and with minimal settings at that.  

Other plugins I've tried might take a stab at the hiss, but they rarely identify it well (or ever), not leading to much reduction or too much reduction and muffling the sound. 

My recent workflow was to use Accentize's dxRevivePro on only the high end, using the Natural setting. It does do a great job removing the hiss, however, at a cost. dxRevivePro struggles and falls apart in high-end content, around the 7-10khz area (or sometimes 14k-17khz), especially with any sibilant consonants. It sounds noticeably artificial in a digital way rather than in the classic overprocessed way. 

So I ended up trading on high-end problem for another. Which still isn't ideal! 

Dialogue Isolate manages to banish the hiss without introducing any unwanted side effects – a true win in my book. 

The noise fader was only reduced about -15db (not even halfway down)! There was no noise pumping, and the quality of the voice was not affected. There were a couple of tiny instances when it had trouble with the RF hiss, but I was able to rectify most of it and fell back on dxRevive for those small spots.

My biggest grievance with the previous version of Dialogue Isolate was that it struggled with high-frequency content. When Clarity VX Pro was released, I relied solely on that to reduce high-end noise, then completely switched to dxRevive/Pro when that was released. However, except for some very troubling spots, while mixing a 45 minute show, I barely used it. Which is unreal.

Also, previous versions of Dialogue Isolate’s Best (offline) quality processing was painfully slower than other neural network-based algorithms. It has now caught up to speed with the others, and dare I say, maybe even faster than some? Hard to say, I'd have to time it!

It's new algorithm is definitely a major improvement for my use cases, and I can't wait to dive in deeper with more trickier recordings!

So, we’ve talked about how it works as an audiosuite/offline processing plugin. Now, for my most anticipated stress testing, how does its real-time noise reduction hold up and compare?

If you own my mixing template, you’ll know I sing Acon Digital’s Extract:Dialogue praises as the best real-time noise reduction plugin available. I’ve tried so many real-time noise reduction plugins, and by far, it gives me the most targeted/natural results that eliminate a large amount of noise comparatively.

What Does It Need to Compete?

In order to compete with Extract:Dialogue(E:D), it needs to perform better in 3 areas:

1. Natural Sound - Does it take away the most amount of noise without sacrificing the quality of the dialogue and the introducing artifacts? Can it eliminate/tame hissy dialogue, misc unwanted tonal noise, and hum?

2. Vowel Power - E:D’s pitfall is that it tends to misinterpret those long, drawn-out vowels, like a triumphant "yeaaahhh!" as tonal noise, which has been a nuisance to automate to ensure quality. Also struggles with breaths and laughter. Does DX Iso fall to the same fate?

3. Group Effort - Overlapping dialogue or group conversations can easily overwhelm and confuse these noise reducers as a cacophony of noise.

What are the Results?

It wins for the interview bus, about the same for the dialogue bus.

How It Wins for Interviews

Some backstory for interview mixing process. Because interviews are typically shot indoors, and boom mics, which sound nicer, tend to pick up more of the room reflections, I have to introduce some de-verbing into the clean up process to tame those reflections. 

With E:D in my interview routing folder processing chain, I can target heavier noise reduction settings to tackle static noise from the various shoots. E:D tames just a bit of the reverb, but does so only because it gets folded into taming noise, which is not their desired effect.

I do own Acon Digital's De-Verberate 3, which is probably the best reverb reducer on the market, however, I don't use it if Extract:Dialogue is being used in my summing bus. 

The reason? It sounds too "processed" and unfavorably gated and soulless when pairing the two together, in my opinion. Reducing the amount either does not enough or too much. 

But I do like RX's Dialogue De-Reverb. It creates a natural reduction of reverb, to my ears, that doesn't take away from the quality of the recording. 

Here's the issue: Dialogue De-Reverb, for whatever reason, is only available in the Izotope RX Spectral Editor. Meaning I have to send the clips to RX via RX Connect in order to process a singular clip. It's wildly inconvenient as I have to process each clip at a time, versus simply audiosuiting all my interview edits in one go. 

It's a time suck, but it does make it more pleasant of a listen, so I give in to that workflow.

But I don't have to waste any more time with RX 11 Dialogue Isolate. Because of the reverb parameter, I can reduce the reverb in real time, and it leaves pleasant results with no tweaking. It literally shaves an hour or two off having to process each interview bite. It saves even more if you're working on an interview-heavy doc!

From here on out, this is absolutely replacing Extract:Dialogue on my Interview Processing Chain. The noise reduction sounds great as well. I find that -12db for reverb and -17db for noise, with a 5.0 sensitivity leaves very little to be tweaked. I can't believe how much time I will save mixing unscripted content now!!

How it Falls Short for Production Dialogue

However, I can't celebrate when it comes to my dialogue summing chain. It holds up to Acon, however it still falls victim to what I had to automate for E:D: elongated vowels and group chats/overlapping dialogue. 

I do have to say it does a better job of affecting those issues. It acts more as a level suppressor than introducing some weird artifacts that E:D creates; however, there are moments when it still does that. 

At this present time, I can find better sweet spot settings with Acon's algorithm that sound more natural than RX 11's. This may change the more I mix with Dialogue Isolate; however, at the moment, I'm hesitant to switch it out just yet. There are a few more considerations to make.

It does do better with breaths and laughter though, which has me automating much less. I would be really happy with this since I did automate much less throughout the show, however, this drawback gives me pause…

CPU Usage and Potential Gremlins

After some quick testing, Dialogue Isolate does use more CPU power than Extract Dialogue. When replacing both instances on the interview and dialogue routing folder in my template, the idle CPU jumped from ~ 39% to 45%. 

This could be a hurdle for mixers working with humbler machines. My system can thankfully handle it, but for those who are already at capacity of plugin processing afforded, this is not the switch to make until your next upgrade.

Additionally, while in an in-person mix review, I encountered playback crashing because of the large number of resources being used from Dialogue Isolate (running native on M1 Silicon). Obviously, this hurts your client's experience with the show by having a crash interrupt. I haven't had this issue with Extract:Dialogue (video engine, another story lol).

When bouncing an offline mix, I found that it took my average 4.6x speed and reduced it to 2.6x. Yikes, running pretty slow as is, and doubles my export time!

However, it did save me an hour or two from editing/mixing interviews, so I'm inclined to let this go for an extra 5 or so minutes bouncing. I usually take that time for a break anyway. However, if you're under a tight deadline, this might still be important to you. 

The Verdict: A Workflow Game-Changer

I think for now, I'm going to keep one instance of Dialogue Isolate for my interviews, and Extract:Dialogue for my production dialogue. Who knows, the more I tweak this may be subject to change, but I feel it's having the best of both worlds. I mixed another episode this way, and no resource crash interrupt!

So does that mean you should upgrade to RX 11? I'd say for now, depending which version you're on, hold off. If you're on RX 7 Standard and below, I think it's healthy to do an upgrade to Standard, as it now includes Dialogue Isolate. 

However, if you're on RX 10 and have another real-time noise reduction plugin you adore, there is no need to upgrade. Hold off til more improvements are made in RX that make it worthy of an upgrade. I fear that may not occur til a few years from now. 

If you don't have any noise reduction tools and are wondering which one to invest in, I'd say go with RX 11 Standard and wait for a deal to upgrade to Advanced if/when you want access to more tools. It's still the best offer that enables you to have many other tools necessary for audio restoration. 

Have you upgraded? What are your thoughts on RX 11?