We show how iZotope's popular restoration package can deliver cleaner audio for your music projects.
Sound is noisy. There's no getting away from it. Whether it's ambient distraction on a location recording, a poorly grounded electric guitar or something unexpected that's been captured in a library sample, the sound we want is all too often mixed up with noises we don't want.
In my experience, taking away unwanted noise is something that is often overlooked in music production. However, it can make a vast difference to the finished product, and save the mix engineer a lot of work. In this article I'm going to use real-world examples to try to shine some light on the under-appreciated dark art of noise reduction.
This first became an issue for me when I moved into a new studio. The live room was a masterpiece of studio design, but the control room had a problem: no matter what I did, I got the classic 50Hz mains hum if I tried to record a guitar in there. It wasn't bad enough to make recordings unusable, but it was enough to bother me. So what can we do when the producer loves the solo and the artist is becoming all gushy about it, but there's hum all over it? There are numerous effective noise-reduction packages, but iZotope's RX is one of the most popular and comprehensive, and it has become my weapon of choice. Some of its processing modules are available as real-time plug‑ins, but in this article I'll be working with the standalone RX Audio Editor program.
This has two main ways of visualising audio: a conventional waveform display and a spectrogram view which represents frequency on the vertical axis and time on the horizontal axis, with colouring used to indicate the intensity of the sound. A slider in the lower-left corner allows you to blend the waveform and spectral views; I have mine set to 100-percent spectral, but it can be very useful to have the waveform showing too.
In a paradigm that's not unlike Photoshop and other image-editing software, RX processing is applied by first selecting an area within the spectral view, and then choosing a processing module within the right-hand panel. The chosen processing module is not active until you 'preview' or 'compare' your settings. Assuming you've selected a region to work on, you can use the 'selection playback' button to just play that region, which can be very useful. My preference is to preview my changes and then when I feel like I'm on the right path, I will use the compare function to be able to quickly A/B the changes. It is surprising how often a proper A/B comparison changes your mind.
Perhaps the most important specific thing to remember about RX is that if you use a module, close it and then reopen it, your settings won't automatically be remembered. So, if you've changed the EQ, for instance, and then want to go back and look what settings you've used, you need to save the settings as a preset using the 'add preset' option. It is also very important to save an RX Document as soon as you've brought the file into the software. If you don't do this, you won't be able to use the history function to go back to any point during your processing.
Taking away unwanted noise is something that is often overlooked in music production. However, it can make a vast difference to the finished product.
Once you've loaded the offending audio into RX, the first step is to consider what you are actually hoping to achieve.
- What are you removing? The answer might sound obvious to start with, but it's often the case that once you listen to something in microscopic detail and with a hypercritical ear, you will find other sounds in there that you question. Decide what your objective is before you start the process. Don't be dragged off course, otherwise you'll go meandering down a path and end up causing more damage than good. If those other sounds you hadn't initially spotted turn out to be a problem at the mix, you can always deal with them in another pass.
- How much do you need to remove? The key word here is 'need'. You might want to remove every hint of noise, but the more drastic your treatment of the sound, the more artefacts you can introduce. There is always a trade-off.
- What does the producer or artist want or need? You might have recorded the part, but unless it's for your own music then you're not in charge.
Once you've made these decisions you can start considering the audio in more detail. I start by working through and making notes on things. It's vital to have good, reliable monitoring for this sort of work, and personally, I strongly prefer high-quality headphones. You may also need a good-quality headphone amp to be able to listen for very extreme treatment, when creating sampled instruments for instance.
Before we start, though, let's take a moment to understand a few simple but important factors that can be overlooked. The first is possibly the most obvious, but in my experience of repairing audio for people it is also the most common misconception: you can't make something that is bad musically better by removing unwanted noise. Surprisingly often, when an artist sends something to me to have noise or clipping removed, they're less pleased with the part once they get it back, because a cleaner recording shows up bad playing more clearly.
Second, remember that noise isn't just one thing. Wanted sound has a 'noise' component, and when it comes to unwanted noise, there is a difference between environmental noise and system noise. System noise, such as tape or preamp hiss, is typically constant, whereas environmental noise is less so.
The last point to consider is that if you are taking away ambient noise (particularly hum and room noise), it's extremely helpful to have a sample that just includes this noise and no wanted audio to use as a noise 'fingerprint'. So, if you have the option, it's better to work on an unedited audio recording that still includes sections where no wanted signal is happening. And if you're making recordings yourself and you're aware that noise reduction might be needed, it's a good idea to deliberately capture such a recording; I try to grab 20-30 seconds of 'pure' noise.
Let's start with a straightforward example from a session I did for an artist creating an '80s pop-style record. As you can see from Screen 1 (above), there are several guitar parts. Some of them are clean sounds, so noise is not being introduced or exaggerated by distortion or effects. Nevertheless, as you can hear in the audio examples that accompany this article, there is still a slight hiss.
The question is whether this hiss can be removed without introducing any artefacts, and whether leaving it there will cause any issues for the mix engineer. Assuming I decide to remove it, I can employ a choice of tools within RX, but my first stop will be to use the Deconstruct tool to break up the signal into its tonal and noise components.
Next I'll look at what I can achieve with just the EQ. It might be that most of the noise is in a specific frequency region, and that if the wanted audio doesn't really occupy the same area, a simple low- or high-pass filter will significantly reduce the problem.
After that, I'll use the De‑hum module (see Screen 2). Even in a basic preset mode, this will do a good job of sorting out simple, low-level hums or hisses.
The clean parts are thus easily dealt with, but they're also relatively unproblematic. The trouble comes when we hit bar 57 and the 'high gain chords' part kicks in, complete with a lot of hiss and hum that's going to cause the mix engineer issues. Screen 3 shows the stereo part loaded into RX, with the left channel on top. You can see a number of 'bars' running horizontally: these are the fundamental and overtone frequencies that make up a note or chord. RX makes it easy to select each harmonic for independent treatment if you want (Screen 4): the wand tool will automatically select notes, or you can manually draw around them if needed.
In this case, though, the problem isn't within the notes themselves, it's with residual noise around them, so I'm going to break the noise down and see what I can remove without causing too much of a problem with the bits we do want. My first port of call is the Spectral De‑noise module (Screen 5). This is a hugely powerful tool, and achieving the best results can take a lot of patience and trial and error. Use the compare option to check things. The 'learn' function is very useful, especially if you have a good section of raw noise to work from. It's important to open the 'advanced settings' option, otherwise you're not using the tools to their full ability.
In this case, the hiss and hum can be suppressed quite effectively to within an acceptable limit by using the 'Hiss Control' preset. The trade-off here is that removing hiss can also dull the top end of the wanted audio. In this case, setting the quality control to 'best' allows me to up the reduction amount for a touch more removal, but the top has still been affected a little so, after rendering it, I will use EQ to add back a little of that lost sound. Once you've learned to understand what you're viewing in RX you can see the changes, but more importantly you can hear them!
Screens 6A and 6B show 'before and after' views of the Spectral De‑noise processing with SoundCloud MP3 audio examples of each. You can see/hear that the processing has removed the high-frequency noise and that, as a result, the audio we want actually looks a little more defined. As you can tell from the audio examples, you can clearly hear the difference.
This was a simple example to get us started. Next time, in Part 2, we'll tackle something a bit more challenging!