So I had this issue last week -- one I could not resolve to save my life. It was driving me insane (which is a short drive to begin with). I advertised the issue on facebook and got some incredible answers. One of which surprised me.
I was answered by a wonderful VO by the name of Shawn McElroy. He was someone that I had seen a few times, but never engaged directly. Man, am I glad I did this time.
We got together on a Zoom call and I have to tell you, I was amazed at the techniques Shawn had imparted to me. Not only did his techniques, which I will go into shortly, resolve the issue, but were so important that I am able to carry them further. Its the "teach a man to fish" scenario.
While talking with Shawn and learning from him what I needed to do to come to my resolution, he mentioned that he uses iZotope RX 9 (full version). Now, I have RX 8 Elements (Suite) which I bought on sale for I think $49. RX 8 Elements alone is $129. Right now, its $199. Its going to change frequently, so I would watch out for it or keep up with it on facebook. Occasionally someone will post the sales.
The only thing you need currently or to start with is Elements. I mean, there are more things you can get with the full version, but 90% of what I needed I could get from Elements.
That said, Shawn also mentioned that he uses it as his DAW (Digital Audio Workstation). I mean, nothing else. No Audition, Reaper, Audacity, etc. I was impressed by that to begin with. A little scary as it is destructive as a standalone program/app, but hey, whatever floats it right?
Now, I know a lot of VOs use RX as their De-click or De-ess software and it does a fantastic job. Usually, you use it with Repair Assistant which analyzes the audio and finds clipping, clicks, hums and noises. You can also use the Repair and Utility options alone or build a Module Chain (which I have not done, but will).
All that said, my main issue was how to remove clicks (mouth clicks or noises) in the words effectively without altering other parts. I had a 10 second clip as a test* that had the click, but when I ran the Repair Assistant or even De-click alone, it would alter another part of the recording. Not Good. If I could hear it, my customer certainly would (he heard WAY more clicks that I did initially).
Shawn showed me some tricks including making a selection within what shows up in RX and doing edits on that segment. One thing RX will do is show a comparison of "Original audio" to "De-click: Settings 1" or so on. This will show you that before/after, before you "Render" the final repair. You have to Render the Repair or Utility action before any changes are effected.
Man, saved my biscuits. Not only that, he went further to show me what a click would look like based on who the recording was made by. Keep in mind, each person has their own set of vocal frequencies including clicks. They show up in various places. Now, before I get into some techniques, let me say this...
The Spectrogram Is Your Friend
If you don't know what a spectrogram is or how to read one...that's ok. Its only the most important thing you can know and if you don't know by now, then shame on you.
Just kidding. Its really is an advanced engineering technique and most really don't know how to use it. But I do, so keep up.
Spectrograms are like an X-ray for audio recordings. Seriously. There are things you can see and affect on a spectrogram that you can't using just the waveform (the squiggly ups and down you see when you record). Most all good DAWs or audio editors from Audacity to Audition will have some spectrogram settings. If not, don't use that software.
Anyhow, here is a typical spectrogram:
You'll see that it looks like something is on fire, but really its an interpretation of the frequencies that it sees. You can see the intensity in -dB and the frequencies in Hz on the right of the yellow spectro. Keep this in mind as it is VERY important in locating and knowing how to remedy various issues.
And here is the waveform alone as well as both together (RX has the ability to show one, the other, both or a variance of either.
If you look at the bottom left above the "00:00:00.000", you will see a slider that will vary between the wave and spectro.
VERY useful for seeing one or the other. I stick with the spectrogram view because I can see what I want easily.
So You Mentioned An X-ray?
Yes. I did. This is the "its going to take some practice" part. As you see or will see on the spectrogram, you will notice it much like a flame that's rising off a fire. Obviously, you know where on the flame its the hottest and that is very similar here. You can see the intensity of the frequency in that part of the wave.
Why is this important? Well, as Shawn showed me, when you can see it somewhere up or down that part of the spectrogram, bingo! You have the issue. Not only that, since it can affect certain frequencies at that point in the waveform, you can surgically edit those parts out. There are special selection tools there on the app that you can use to really get down and dirty with editing in a specialized way.
From the left are the main Selection tools. This tool will select the whole segment of the waveform top to bottom. Then a selection whereby you can select parts of a waveform either in width or height or both. The third one will select a band of frequency across the whole audio. It might be good if you have a hum or noise at a certain frequency and can see it as a line across the spectro. Kinda like using a very narrow band EQ. Be careful with this one. The fourth I haven't really played with but its a lasso tool to select irregular shapes. There is a brush tool that can vary the width of the brush and also be used to 'paint' a selection on. Much like any paintbrush tool can do. There are other tools, but the ones I mentioned are the most common.
What I wound up doing on an issue was to look at the spectrogram and see if I can see an offending pattern or shape. I did and what you will see below is an edit using the tool second from the left:
You can see there is an area where I used the surgical edit to remove some odd buzz. It only affected that part of the frequency and nothing else. No dropped words or blank spots.
Ok, Cool. What else Can It Do?
I'm glad you asked that question. When we read a script or book or whatever, there will inevitably be breaths. You can use any de-breath plugins that you want or have or feel comfortable with, but Shawn also turned me onto a solution in RX that can be used to get rid of those unwanted breaths in between sentences. Those spaces are also breeding grounds for lip smacks and mouth noises. Dreadful. So by using this technique below, you can mitigate them. This involves not deleting or clearing out the spaces to 'blank', but rather filling it with noise, but also consistent noise.
I have a file called RoomNoise.wav. It is what it says: room noise.
What I can do is to select, say 5 seconds of this room noise, copy it and then make a selection in the file I need to edit. I can Edit>Paste Special>To Selection to paste the copied room noise selection into that area (Alt-Shift-V on Windows or I would presume Command-Option-V on Mac. Not sure).
This one shows a breath. Notice the waveform does not even hint at the breath, but you can see it in the spectro.
This is the same one where I plasted in 'room noise' to the offending area.
As you can see, there is an area that does not look like the others, but I guarantee, you will not hear it and by pasting into a selected area, it will not move anything! Great huh.
So Moving On
So moving on now, let me say that there are so many other things I can't even touch on because I haven't even gotten to that yet. More on that later, but I hope this has been a great start to what you can do with iZotope RX and Spectrograms. Keep watching and there should be more.
EDIT: If you want to watch a really great tutorial on Youtube, check this one out.
Comments