Complete Guide to Podcast Audio Clean Up for free
Riley Byrne, Owner, Podigy.co
22 September 2017
Welcome to our second official guide to podcasting! Today we’ll be going through all the ways to clean up audio (for free!) using Reaper plugins. If you missed out on the first guide, our Complete Guide to Podcast Editing, I suggest you take a look through it, because we’ll be building off some of the concepts introduced there.
This guide will show you 4 techniques for reducing noise and giving your listeners a better experience. They are:
- Removing background noise
- Restoring off mic vocals
- Reducing plosives
- Reducing noisy events.
Removing Background Noise
The first step to getting our audio tracks clean is to reduce the amount of background noise. By “background” noise, I mean the noise that is always present in your room. It’s consistent, fairly unwavering and is present throughout the entirety of the track. Common examples are computer fans, AC hums and radiators.
Obviously, the best advice is to try to curtail these problems before you turn on the mics, but unless you’re recording in a studio, every one of your voice tracks will have some amount of background noise in it. So instead of building a custom studio, we’ll use ReaFir to reduce some of this noise.
ReaFir is an incredibly powerful EQ and dynamics processor that uses FFT spectrum analysis. Think of it as an EQ or multiband compressor with a near infinite amount of bands. This allows us to make very minute edits, and I’ve got some plans for some interesting guides in the future. But for now, let’s start reducing the noise in our track.
For this tutorial, I’m going to work with a few pre-recorded audio bits so you can more easily hear the before and after. So we’re starting off with a track that was recorded with an AKG Perception 100 (my very first mic!), with the noise of an air conditioner clearly audible in the background:
No Noise Reduction
We can hear some an unacceptable level of fan noise. So lets go to FX > Cockos > ReaFir (FFT EQ +Dynamics Processor), and add it to our track.
Arrows have returned from the last guide as well.
In the interface for ReaFir, click on the menu next to Edit Mode and choose Precise, FFT Size to 1024 and then underneath change Mode to Subtract. When you select this, you should see an option appear marked Automatically build noise profile (enable during noise), click it. We’ve now primed ReaFir to to listen to the noise in our track, which we’ll select in a second. While Automatically build noise profile (enable during noise) is selected, make sure you only let Reafir listen to static noise, otherwise it will start to incorporate your voice into the noise profile, and try to eliminate it from the track.
At this point, you may want to save this as a preset for ReaFir by clicking the + button next to Param and clicking Save Preset, or Save Preset as Default, if you want this to come up automatically when you load an instance of ReaFir.
Now, find a piece of your audio where there is only background noise. Make sure no one is talking, there isn’t a sniff or cough, stifled yawn or beer can opening. Right click to highlight this section of audio and click the Repeat button next to the red Record button at the top of the page. Set your play cursor somewhere within the time selection and hit play.
Normally you can eyeball a spot to build a noise profile but listen through it a few times to make sure there isn’t another sound in there that could throw off ReaFir’s calculations.
You’ll notice the red line in the GUI of ReaFir start adapting to the noise. Let the audio loop until the red line has just about settled down and then deselect Automatically build noise profile (enable during noise) in ReaFir.
Now when you play back your audio, you should hear it noise free:
Noise Reduction One Pass
However, you’ll also notice that there are some artifacts in our audio that weren’t there before. The noise is gone, but the voice doesn’t sound natural. To correct this, CMD + Click (Ctrl + Click on Windows) the red line and start to drag it down. You’ll see the line retains the shape of the noise profile, but dragging it up and down will adjust the amount of effect it has on the audio. Pull it down until the voice sounds normal again, even if some noise is slipping through. Like many aspects of audio editing, the trick to getting the best denoising results is to use a few instances of ReaFir each taking off some noise rather than one instance trying to remove it all.
Add a second instance of Reafir and make sure it comes after the first one we set up. Repeat the process of creating the noise profile for this second ReaFir, but this time it is only working on eliminating the noise the first one is letting through. Once again, listen to the resulting audio and pull down the second ReaFir’s noise curve until it is not adding any artifacts to the voice. While often 2 passes of ReaFir will bring the noise level down to an acceptable level, don’t be afraid to use 3 or even 4 passes if you’re going for a really pristine sound. However, I’ve found that after 4 instances of ReaFir you’re probably doing more harm than good to the signal, but in the case of our AC addled example a third pass was warranted. Listen to the original, one instance of Reafir eliminating all the noise, and three instances of Reafir eliminating a bit of noise each, below:
No Noise Reduction
Noise Reduction One Pass
Noise Reduction 3 Passes
Note: Multiple instances of ReaFir across multiple tracks can get a little CPU taxing, so you may want to render out the audio by selecting all the tracks (not the audio, the track parameters itself on the left hand side) and right clicking to select Render/Freeze Tracks > Render tracks to mono stem tracks (and mute originals). This will free up processing power for other effects down the line.
Restoring Off-Mic Vocals
Next let’s look at a less conventional way of using ReaFir for when people are sharing mics, or someone has said something off mic that we would like to use in the podcast. This method is something I’ve used only when the audio for an interview is so bad, that it is borderline unintelligible. In these cases, I take the track as far as I can using ReaFir, using up to 4 instances on the track, and then create a duplicate of the track, and denoise that one further, with the sole intention of getting out all the noise. This will make the secondary track sound like garbage on its own, but not to worry. We then adjust the level of the heavily denoised track until it is accentuating the voice main track, but not so high that we can make out that there are two separate tracks. Doing this can add the little bit of clarity that can help save a botched recording session, although be warned the results will be far from perfect.
We’ve talked about plosives before, but just a quick recap: plosives are the ‘puh’, ‘buh’, ‘tuh’ and generally any vocal sounds that have a large amount of air expelled, that is then translated into a lot of low end energy by our microphones. Most plosives can be mitigated by using a pop filter/wind screen, but eventually you’ll have to deal with one while you’re editing. Luckily, we can automate the process in Reaper so dealing with them only takes 1 button press.
First, go to Preferences > Appearance > Media and make sure Per-take FX is checked. This will let us apply FX to only certain parts of a track, which is useful for plosives, because we’re going to apply an EQ curve to just the “puh” sound that would not sound good applied to the whole track.
Next, open up a blank track and open the FX tab. Load up an instance of ReaEQ and reset the EQ to factory defaults, and then set the Type of band one to High Pass, the Frequency to 150 Hz, and the Bandwidth (oct) to 2.00. Next, right click on the listing VST: ReaEQ (Cockos), and click on FX Chains > Save Selected FX as Chain. This part is particularly important: Save the FX chain with a name starting with ‘01’ and then a space. Reaper looks at FX chains alphabetically, so we need this one to always come up first.
Finally, go to Actions > Show Action List…, and then click New next to Custom Actions. A new screen will pop up. Give the custom action a name (‘Plosive Fixup’ comes to mind), and then add the actions Item: Split items at time selection and SWS/S&M: Resources – Paste FX chain to selected items, slot 1. Click OK, and then back on the action list screen give the action a short cut. If you’re also using our Complete Guide to Podcast Editing, I suggest ‘D’, as it will keep it close to the other actions we have set up for editing.
From here, it’s just a matter of highlighting the plosives (really make sure you zoom in and only select the “puh” sound), and pressing D. This will add our de-plosive EQ to just that section, leaving the rest of our audio untouched and making our listeners very happy.
Reducing noisy events
This part of the guide is based on brand new features that were just added to Reaper and thus is subject to change the more the community starts using it and developing tools around it. For now, however, let’s look at how Reaper can help us bring down the sound of nonstatic, noisy “events”, such as a car honk, birds squawking, door slamming, or a beer being opened during a podcast. Oftentimes it’s just easier to edit around these happenings, but they’ll inevitably happen at critical moments, and it’s our jobs to make them sound as good as possible regardless.
First, we’re going to go to Options > Peak Display Mode > Peak Display Settings. This should open a new window, and from there we want to change our Display Mode to Spectrogram. You should see the waveform be replaced by a colourful collection of blobs that represent where the energy in our track is. You can play around with all of the settings in the Peaks Display Settings window as much as you want to get a better view of your audio, it isn’t actually affecting the sound and can give you a better look at what you’re trying to find. In this example we have the audio of me saying “I am drinking a beer,” with the sound of a can opening in the background.
No Spectral Processing
Let’s go to our Actions > Show Action List…, again and search for Spectrogram: Add spectral edit to item. In keeping with the theme of using our left hand for key commands, I’ve assigned this action to the ‘A’ key.
Next, let’s look at our spectrogram for any inconsistencies. You’ll see that most of our speech is concentrated in the lower portion of the spectrum and even sibilance doesn’t come up too high. This is great for us, because in this example, the beer can snaps can be clearly identified by how much more of the high end they take up.
We’ll highlight one of those snaps by right clicking and dragging and press ‘A’ to create a new spectral edit box around it.
Here, the white line is the spectral edit box. In the blue box is the main energy of the beer snap sound, although it extends all the way to 22 050 Hz. The knobs in the green circle are our controls for the spectral edit, and the area in the black box is the energy from the spoken word.
These spectral edit boxes are unlike anything we’ve worked with before. You can click and drag them, expand or collapse their range, and they will only affect the area within the box with the parameters you set for it. You’ll notice on the right hand side there are seven knobs that you can bring up. The top one is for gain, you can boost or attenuate the signal within the box. The second row is for time and frequency fade, respectively. These fades allow you to blend your edits more naturally with the audio to the left and right (time) or above and below (frequency) the edit, like the knee of a compressor would. The third row allows you to compress or expand the audio within the box, like a very specific multiband compressor. Finally, the bottom two knobs allow you to set a gate threshold and gate floor. In the future these might be useful for clearing out ambience in a track. But for now, we’re mostly concerned with the gain knob.
Click around the edges of the box to resize it. CMD + Click (Ctrl + Click on Windows) around the edges will change the cursor into a pen, and you’ll be able to draw a custom shape. For our example, we know the ‘snap’ we’re trying to attenuate extends much higher than the voice, but has a relatively short decay time, and most of the energy seems to be concentrated just above where my speech is.
Here we’ve drawn around our voice and tried our best to only reduce the gain of the snap.
This next part is not fun, or pretty. Unlike many concepts of podasting, there are no hard and fast rules for spectral editing, as each case is going to be different. Experiment with the knobs, particularly the gain and fade knobs. Generally use the gain knob to bring down the sound, and the fade knobs to make it sound as natural as possible. It isn’t fun, and the results will seldom be perfect, but in a pinch it can rescue a critical piece of audio. Here is our example after reducing both instances. It isn’t perfect, but it is significantly reduced:
With Spectral Processing
As I said before, the Reaper spectral editing has only been out for a little over a week, so I expect these features to really start to shine once the community starts tinkering with them. If there is a free, universal denoising tool that I’ve missed, please don’t hesitate to drop me a line at firstname.lastname@example.org. I was midway through writing this article when Reaper 5.50 dropped, so I had to revamp just about everything in the last section to reflect this, and I won’t hesitate to do so again if you know of some hidden gem that everyone is sleeping on.
Anyways! Spectral editing, like all denoising techniques, is more about trying to manage a problem as best as possible, not eliminating it. Oftentimes, it’s better to have a more natural sounding podcast with some noise than one with lots of artifacts from over using these tools. There are little errors in all of your favourite podcasts that you’ve consciously or unconsciously forgiven. The Nerdist really seems to hate de-essers, The Flophouse has a remarkable amount of crosstalk, and Flock of Seagals talks exclusively about Steven Seagal films. All are flaws, but we love them anyways because the content is what we’re there for. So keep that in mind while you’re using this guide (and our other guide) and make sure you’re never putting the quest for perfect sounding audio before the goal of entertaining content.
Want more help podcasting?
Leave your name, email and a link to your podcast and Riley will listen to your podcast and make you a personalized action plan to take it to the next level.
Want to Learn more?
Here are a selection of articles to help you become a better podcaster
PODCAST RECORDING The Best Microphone for Podcasts Riley Byrne, Owner, Podigy.co 14 December 2017 iTunes | Google Play | Stitcher So now that we’ve done software, let’s look at microphones. We touched on them briefly in our guide to podcast recording, but today we’re...
PODCAST EDITING Best Podcast Editing Software Riley Byrne, Owner, Podigy.co 06 December 2017 iTunes | Google Play | Stitcher If you’re familiar with Podigy, you probably know that we like Reaper. A whole lot. And if you’ve followed any of our guides, you know that...
PODCAST EDITING Boutiqu Plugins And Podcasters Riley Byrne, Owner, Podigy.co 01 December 2017 iTunes | Google Play | Stitcher Right now we’re in the middle of sale season, having just come off of Cyber Monday and soon enough Boxing Day (for us Canadians) will be upon...
PODCAST EDITING Gain Staging for Podcasts Riley Byrne, Owner, Podigy.co 17 November 2017 Want more great tips? Subscribe to The Podcaster's Podcast: Gain staging is a pretty simple concept. Basically, it’s how you set your volume levels throughout the signal chain to...
PODCAST EDITING Gating for Podcasts Riley Byrne, Owner, Podigy.co 02 November 2017 Gates serve a very specific purpose, and in the context of podcasting I often find that gates are unnecessary and can even be detrimental to the final product. Gates essentially only...
PODCAST RECORDING Bit Depth, Sample Rates and Buffer Size for Podcasts Riley Byrne, Owner, Podigy.co 20 October 2017 Today we’re going to walk through a few things that every podcaster has to decide upon before recording. In our Podcast Recording Guide, we outline...
PODCAST EDITING Master FX for Podcasts Riley Byrne, Owner, Podigy.co 15 September 2017 Google “audio mastering”, and you’ll undoubtedly come across dozens upon dozens of sites that call mastering “the dark art” of production. To a large extent this is hyperbolic, but...
PODCAST EDITING Multiband Compressors for Podcasts Riley Byrne, Owner, Podigy.co 14 September 2017 Following up on our article about de-essers, today we look at multiband compressors. Both are frequency-specific (a la EQ), compressors, but whereas de-essers only...
PODCAST EDITING De-essing for Podcasts Riley Byrne, Owner, Podigy.co 08 September 2017 A de-esser is a podcaster’s best friend. It reduces the sibilance (words with “sss” or “shh” sounds, like ship, chip, sip, zip, or jump) of phrases so that they don’t jump out in...
PODCAST EDITING Compression for Podcasts Riley Byrne, Owner, Podigy.co 06 September 2017 Compressors are perhaps the most misunderstood and misused of all the tools podcasters have at their disposal to edit podcasts. They’ve been on every single episode of a big name...