Pro Audio Files

Free Mix Workshop Premium Courses

Solving 3 Common Audio Post Production Problems with RX 6

Jonathan: We’re going to talk about some of the new modules, the new technology that’s showing up iZotope RX 6, and I think the thing that we’re going to start with is called Dialogue Isolate. This is one of the modules that leans into machine learning technology, and the weight of that expresses itself in this module.

I guess machine learning tools have the ability to help us not so much on the noise that we’re trying to get rid of but rather the signal that we’re trying to preserve. So it really sort of flips the script when we’re doing restoration. So talk to us about Dialogue Isolate.

Mike: Yeah, so Dialogue Isolate, you’ve hit the nail on the head, rather than trying to remove or reduce noise here, we’re actually trying to say, “This is the bit that we want, we want to isolate, we want to extract the dialogue from a bed.”

Now, this is one of my classic audio restoration test files. It’s a recording from the trade show floor, and the reason I chose this for Dialogue Isolate, is essentially, it’s Dialogue over Dialogue, so Dialogue Isolate knows the algorithm — the machine learning algorithms know what dialogue looks like, but hang on, the unwanted signal is dialogue as well, so would it do it?

So let me just play the example before we start processing it.


So typical trade show recording. So we go into the dialogue isolate module, and if I just take this sort of increase intelligibility preset, you’ll notice here that with the machine learning options, there is not preview option. So we need to in essence, either use the Compare button, or in this case, for ease, I’m going to simply hit the process button, and it just takes a few moments to process through.

Jonathan: You always have undo, right?

Mike: Absolutely. And now if we play that — so if I just play this from the beginning here…


So it’s really helped to bring the dialogue — the background noise is still there, so it still recognized it was a trade show floor, so it’s still believable, and so we’ve matched the pictures, but of course, there are scenarios where we might want to change — especially in my world of radio where I don’t have the pictures to worry about. The pictures of course are better on radio of course, but let’s not go there, so I might want to change the environment altogether, and so what we can do is we can go for an all out where we go to in this case, the separate dialogue.

So if I just go back to the initial state, and we pre-process that, what you’ll find is quite an amazing end result.

So you can already see from the spectrogram how much cleaner it is, and if we just play it, in essence…


Jonathan: If you were going to mix this narration now into some other background, you wouldn’t notice the fact that there’s something gone. There’s something missing.

Mike: So it enables me, especially in radio, to change the location of the recording.

Jonathan: Is that okay to change the location? I guess it is. It’s an illusion, right?

Mike: Yes! [laughs] So that’s, for me, that was a real testament to the coding work that you guys here at iZotope are able to potentially take an algorithm, which is dialogue over dialogue, and still pull out what I wanted, and get rid of what I didn’t want.

Jonathan: Fantastic.

So now we’re going to go into another module called De-Rustle. We have an engineer here by the name of Russel, and this feature, when we were thinking about naming this feature, he became very concerned. He thought it was a pink slip, in fact, it turns out to be job security, because Russel is kind of awesome.

So tell us a little bit about what is rustle?

Mike: So more and more TV Drama is now having to be recorded with personal radio mics on, because the director is wanting to shoot two or three camera angles at the same time, so the conventional boom pole, which we all would prefer to use, is no longer viable, because we can’t get the boom close enough, because of the wide shots, and of course, as soon as you put these sorts of microphones on, especially in drama where they have to be hidden so that they’re under the clothing, we get problems of the mic either picking up the clothing moving, or worse still, the clothing actually moving over the microphone and getting this rustle effect.

The problem with that is conventional restoration technology failed, because the rustle is similar frequencies to the speech that we want, and it’s often at a similar audio level, so either of the two techniques that we would conventionally use don’t work. So it’s always been a major challenge to get rid of the rustle — the clothes rustle from recordings of personal radio mics.

Jonathan: Right. So let’s have a look.

Mike: So this is again the deliberately setup demo file to prove it, because more often than not, the material that I use for clients, I can’t of course use in these sorts of environments, for clearance point of view.



So all the classic issues of rustling clothes that we would have to deal with, and so again, this is a machine learning module, so no preview button. We have two simple controls, reduction strength and ambience preservation. So again, as I always do, the default settings are often really helpful, so I’m going to simply process this and we’ll hear what happens.

[dialogue, de-rustled]

Now what you’ll probably be able to hear, is especially in this long section through here that the ambience level will have just dropped a bit, because in trying to deal with the rustle and reduce the rustle sound, it actually tends to push the ambience down. Of course, so you essentially start getting a pumping of the ambience, and this is where the ambience preservation control works really well.

So if I just go back to the initial state, and we push the ambience preservation all the way up, it’s still going to get rid of the rustle, but the ambience won’t start to pump, especially through the quiet section.

Jonathan: You can even see it there in the spectrogram.

Mike: Yeah, you can see the classic sort of muddled effect. So again, let’s just play that.

[dialogue, de-rustled]

So there we have it. We still get rid of all of the rustling that was going on, both in this big gap, and also underneath my speech, but the ambience stays at a natural level.

Jonathan: That’s great. This is actually a real life saver, this actually saved me, Nate, and Matt, who are video producers here at iZotope, are especially grateful, because I put on my own lav for one of my episodes of Headroom, and there’s a saying about never let a mastering engineer put on a lav mic. Anyway, I’ll never do that again, but this saved the episode of Headroom, so yeah.

So now we’ve gotten rid of rustle, although we love Russel, and —

Mike: You have any employees called Wind? [laughs]

Jonathan: Uh, Wendy… No, but I have tried to make recordings on a windy day! So yeah, sometimes the weather doesn’t cooperate, so let’s talk about de-wind.

Mike: So yeah, wind noise, whether it’s nature moving wind around, or often in studios now, there’s a greater use of fans to push the smoke and the hair, and this that and all the rest of it, so we have essentially wind is air movement over the diaphragm of the microphone, interfering with the microphone working properly. And you can see on this example it’s really clear in the spectrogram, because essentially, with the natural default colors, the brighter the yellow, the brighter the amber, the louder it is, and in fact, you can see on the audio waveform here where we get a huge amount of wind.

So let’s just play that.

[dialogue, windy]

So diaphragm is really getting quite battered around in this case. So again, we can go for the de-wind module, and we have a range of settings, so if we use this remove rumble, because obviously it’s lots of low frequency, so we’ve got the spectrogram showing here on extended log so we can see all of the low frequencies nice and clearly.

So let’s just preview this one.

[dialogue, de-wind]

You can start to hear some of the higher frequencies that it hasn’t quite got. So what I’m going to do now is to use a preset. Here’s one I prepared earlier, but the critical thing is what I’ve done is I’ve increased the fundamental recovery. Now, what that’s going to do is that’s going to bring back some of the lower frequencies of my voice, but I’ve also upped the smoothing and also, more importantly, upped the crossover frequency, because in the remove rumble preset, the crossover frequency — in other words, the frequency above which it won’t do any processing is about 200Hz. So that’s why we were hearing some of the high frequency wind elements left behind.

So by increasing the crossover frequency, increasing the fundamental recovery, now when we hit preview, it gets rid of the wind, but also my voice is natural.

[dialogue, de-wind]

And so even here, where there was a lot of activity, and in fact, probably the diaphragm was starting to hit the end stops with the pressure from the wind, we’ve been able to recover the wanted speech audio. So again, another absolutely brilliant life saver.

Jonathan: Great. So while this is called de-wind for instance, I can imagine if it was thunder in the distance, that is to say some low — very low frequency signal that’s still inter-modulating with the voice, that it would be applicable. Do you have any tips or guidelines to when you’d use one module or the other, or is it sort of a matter of trial and error?

Mike: It is, and it’s interesting, it’s a matter of just sometimes, thinking outside the box, so there have been occasions where I’ve had a file which has got a huge amount of hum and buzz on it, so in fact, some of the harmonics are beyond the 8 harmonics that de-hum module allows me, and actually, I used Spectral De-noiser and the blue line, and being able to focus in and process more heavily certain frequencies. So again, it is a certain amount of experimentation of course, the more you use any tool, the better your sense of what the different elements can and can’t do, but no, absolutely, don’t look at the label and think that’s all it can do.

Jonathan: Well thank you very much. This is a really great explanation of some of these technologies, and we hope you all find them useful. Thank you Mike Thornton from Pro Tools Expert. I’m Jonathan Weiner, and thanks for joining us.




iZotope develops award-winning audio software and plugins for mixing, mastering, restoration and more.

Free Video on Mixing Low End

Download a FREE 40-minute tutorial from Matthew Weiss on mixing low end.

Powered by ConvertKit
/> /> /> /> /> /> /> /> /> />