Eddie Bazil:  Hello and welcome to the Sound On Sound Recording & Mixing podcast channel. I am Eddie Bazil. In this podcast, I'm going to share some tips on how to apply specific corrective processes to vocals, in particular band-pass cleaning, de-essing and removing resonances and then I'm going to show you how to manage vocal lines within a mixed context. Traditionally in the pre-botox days, one of the roles of the mix engineer was to clean and optimise the recorded audio files and then to perform the level and pan mix cleaning involved, removing anomalies from recordings and optimising dealt with managing dynamics. Leveling referred to gain staging and panning, referred to the placement of sounds within the stereo panorama, which in turn helped to resolve the three nasty problems faced with all mixes - that of summing, clashing and masking of frequencies. In the first example, we're going to concentrate on cleaning and in particular, how to address one of the two obstacles all track vocal recordings face - clicks, implosives and sibilance. We'll concentrate on sibilance for now. These anomalies occur either because of the vocalist delivery or a side effects of the hardware used, microphones being a classic example. Clicks can easily be removed with declicker software or by manually redrawing or erasing each click using an audio editor. Plosives are heavy breathy consonance, which when sung, direct the burst of air towards the microphone's capsule resulting in an audible pop, b, k and p being prime examples and there are specific tools that can remove plosives without much hassle. SSLs Vocal Strip two has a useful diplo, which ranges from 20 to 250 hertz. This is where most plosives lie and is a fast and effective way to remove pronounced explosives. However, manual micro editing can yield better results. Finally, we have sibilance. Sibilance refers to the harsh hissing sounds produced by certain consonant sounds like s, f, and sh. I'm leaving mic stand bumps, sniffling, coughing and local earthquakes out of this equation. I'll start by de-essing, a simple female vocal line written by David Plummer and sung by Becky Bremner. The weapon of choice here is FabFilters Pro-DS de-esser. The first step in all corrective processing is to clean each audio track by using band-pass equalization. The aim is to remove redundant frequencies that we cannot hear and to leave what we refer to as the suite frequencies for the ensuing production processes. With tracked audio, there are almost always redundant frequencies to remove, some examples being low inaudible frequencies which can clutter and darken reverbs and quite often false trigger compressors due to the high amplitudes at the other end of the spectrum, high inaudible frequencies can make reverbs sound brittle and coarse and make coloured equalisers sound harsher and hinder any airband processing. The airband refers to a high frequency range that when boosted, adds clarity, brightness and a sense of spaciousness to a sound or a track. Okay, let's start with the first example. First, the unprocessed version. And now the processed version to tempted by the prize still up. Dynamic Equalisers are the new faux compressors of our times. They're far more versatile than static equalisers and are perfect for both corrective and colouring tasks. A dynamic equaliser functions much like a static equaliser, but adopts certain compressor functions, most notably threshold and attack and release. A dynamic equaliser applies the game change directly to the gain parameters of a multi-band parametric equaliser. As with most dynamic processes, the threshold determines at which point gain changes take place, be they compression or expansion, or both. You have control over the bandwidth denoted by the Q value and much like a compressor, the compression expansion response and behavior is controlled with the attack and release functions. Think of dynamic equalisers as multi-band compressors on steroids. However, there is a distinction between the two topologies - whereas multi-band compressors use crossovers, dynamic equalisers utilise all the traditional filter shapes along with bandwidth control and these simple distinctions give the latter more precision and control over the audio being processed. I've used fab filters Pro Q3 to band-pass the vocals. Q3 is a dynamic equaliser that provides both compression and expansion, but without a ratio attack and release functions. It is my go-to equaliser for all manner of corrective processing. A high-pass at 48Hz using a 24dB slope and a low pass at 12.2kHz using the 24dB slope has removed all the redundant frequencies. This ensures none of the offending frequencies will hinder any further processing and in particular, the de-esser process. Although the ance is not aggressive, it pays to attenuate gently to afford a delivery that doesn't sound like it's been chopped to shreds. Always keep in mind that words are being sung and they need to be kept true. Our job is to manage extremes and errors, not to rewrite the alphabet using only vowels. FabFilters Pro-DS is a simple, yet powerful de-esser and it can serve a whole host of functions that are not exclusive to taming sibilance. I love using de-essers to smooth out reverb artifacts or to use the middle and side feature provided by some de-essers to master mixes rather than single vocal tracks. More on that in a future podcast. Let's concentrate on the settings I've dialed in. Threshold determines when the de-esser triggers. It is gain dependent and not frequency dependent, much like a compressor range scales the detected gain reduction so that it stays within a desired range. Use your ears to tune this setting. Although we engineers love numbers, I find that once you have the basic setting sorted out, the ears become the most important weapon of choice to determine overall texture, shape and presentation. The high and low pass IE band pass side chain filters come with a very useful audition feature, which allows you to hone in on the exact range that the sibilance lies in. Single and all round modes are self-explanatory. If you're processing vocals, then single is where you start as this will split the sibilance from the nonsense. All round is very useful for entire mixes. Wideband and split band processing comes next when wideband processing is enabled, Pro-DS will lower the overall gain of the audio when sibilance is detected. When split band is chosen, only high frequencies will be attenuated. Again, use your ears to determine which works best for the given task. I've used split band. Look ahead is very useful as it allows the de-essing to start before the trigger audio level actually exceeds the threshold transient and starts of sibilant consonants can be grabbed early and in their entirety. I've set the threshold at -34 dbs, the side chain filter from 5-12k as Becky's sibilance sits in this range. I selected 6dB for range as I wanted to keep the attenuation gentle and smooth, and look ahead set to 15 milliseconds, the maximum allowed. The result is a non-aggressive and smooth attenuation of sibilance without taking away from Becky's delivery. Some sibilance is required, so don't overcompensate the processing, keep it natural. In the next example, I'm going to use an equaliser to tame vocal resonances. Resonant vocals can sound painful and can ruin a great vocal delivery. Causes range from types of microphones used and how they react to the voice, the space and room that the vocalist is recorded in and the texture of the vocal delivery, be it the qualities of the voice itself or how the vocalist sings. We've all heard vocal resonances and they're so common that we've created software designed to handle this very specific anomaly. Let's run through two ways of dealing with vocal resonance. The first involves using a static equaliser to locate, isolate and tame the resonances and the second in entails using a dynamic equaliser that has a great feature that offers an automated detection tool to isolate and dynamically tempered resonances. I'm using the vocals from the puppet track, sung by Yvonne McEwen. First, the dry version. And now the static equalised resonate version. And again, but with auto resonate. There is a trick we poor, poor engineers use to locate residences. It is a simple trick from a time when life was simpler, the golden years whereby Nike stuck to designing trainers and not eyebrows and empathy was a proud word in our vocabulary. Ah yes, them days. So the trick involves creating a single bell-shaped EQ note with a slope of 12dB, narrowing the bandwidth or Q, but not too narrow. We don't want to notch, not yet. Anyway, boosting it by 4-6dB and then scanning and sweeping the audio's frequency spectrum from low to high. This technique will locate and exaggerate any resonances. Once you've detected a resonance, start to narrow the bandwidth until it's tight enough to envelope only the resonance and not the nice frequencies around it. Once you focus in on the resonance, grab the nodes, gain knob, slider and reduce it until the resonance is attenuated but not completely removed - we don't want to cut into the vocals as very aggressive resonance removal can make a vocal delivery sound unnatural and take away from the vocals overall texture. Continue until all resonances obtained. Assuming there's more than one, I find FabFilter's Q3 an excellent tool for the job, not just because it's a very well coded equaliser, but because its winning Gooey makes the job of finding frequency anomalies easy. I'm a firm believer in using every arsenal at my mixing disposal. I don't just rely on my dodgy ears, a visual guide can be as potent as hearing. I ended up locating two resonances that were disrupting the vocals, one at 322Hz and another at 770Hz. I created simple bell-shaped nodes for each. Both nodes have very similar Q values around 8, and both have been attenuated by -7 and -5dBs respectively. This is static processing and I made a conscious decision to use it as such to show you how effective and simple it is to nail resonances and how natural the results are when attenuating and not extreme cutting. Q3 can also be used as a dynamic equaliser and normally I would use this mode to process resonances as the results are always more natural than static cuts but I'll show you this process with another equaliser, Tokyo Dawn Labs Nova GE Equaliser. Nova GE has some very useful tools to help the user achieve various equalisation tasks, be they corrective or creative and one of those tools is Smart Ops. This tool analyses the incoming audio, creates an EQ response and then offers a bunch of processes to process the response. You can match the response using a reference file or the built-in pink noise generator. Operations include creating a starting point EQ template with selected EQ notes, multi-band processing and dere, both statically and dynamically. In this example I'm using dynamic de-resonate. You hit the learn button, play the audio you want to process, stop the process. Once the whole audio file has been analysed and Nova creates an EQ response of the audio, you can then select the reference file or pink noise to apply any of the chosen processes on offer. I chose pink noise as the reference for this vocal line and Nova found resonances right across the audio. It did this because I used a wide range between 30Hz and 10kHz and therefore it searched for ressies within that range, adjust range to encompass the frequency range you want to process. Once all the EQ notes have been created, you can switch off bands that you don't need and then fine tune the bands that you want to work with. I've disabled all the bands bar the three I want to use. The first is set to 330Hz, the second at 740Hz and the final band at 1.6kHz. I've used conservative threshold and ratio settings as I don't want any dramatic dips. Once you're happy with the result, use apply to commit the process. The advantage of using a dynamic equaliser for removing resonances is that once the resonances have been attenuated, the treated bands can jump back up to their normal state as determined by the compressor's release value. This affords a more natural response than a static cut iLIGHT. The results of the above processes and as always these processes are context driven, so you use your ears to determine which process affords the best result. In the next example, I'm going to cover an old and trusted technique used to duck a vocal reverb to allow the starts of sung words to play through dry and affected by the reverb trigger. In a mixed context, it can be very easy to fall into the abyss of clutter and smeared textures. It's bad enough trying to achieve natural separation between multiple tracks, let alone managing the clutter and smearing of effects and the biggest culprit is reverb. Reverb by its very nature is a collection of reflections that result in a smeared response. If you solo the reverbs in your mix, you will be met with a wall of smeared frequencies with vocals. This can be a nightmare as we strive to maintain clarity, but want a need to run it through effects for spacious depth or to denote a given space. In a mixed context, every time a vocal line is delivered, the reverb is activated, irrespective of the duration and dynamics of each song. To understand this a bit better, let's remove the obstacles of nailing the right reverb settings for a vocal line and concentrate on what happens when a vocal reverb is triggered. The minute a sound is sent to a reverb, the reverb will activate unless one specific parameter is utilised - that of pre-delay. The pre-delay simply denotes when the reverb is triggered in effect, how long it takes for the reverb to kick in. A suitable pre-delay value will allow the start or onset of some words to be unaffected for the specified pre-delay time value until the reverb kicks in. For most basic vocal reverb tools, this would be enough to guarantee a level of clarity. However, with a reverb that is ethereal and evolving, it can be problematic. The next critical parameter to shape is decay. Decay denotes how long it takes for the reverb tail to dissipate, or how long the reverb lasts for as it fades out. If we have set amounts for both and the reverb will adhere to those values every time it is triggered by a sound, this does not allow us to shape the behavior of these values. The reverb literally kicks in and then fades out irrespective of the duration and dynamics of the sound being fed into it with words. This can be a real headache. Words can vary in length, dynamics and delivery, so having a reverb triggered using the same fixed values can result in some words sounding nice and crisp on the onset, while others will smear in texture and sound lacking in clarity and energy. Using a compressor to duck a reverb is a great way to counteract these pitfalls because the threshold is used to trigger the compressor. To compress the reverb, we can specify which words will trigger the compressor and which will pass through directly to the reverb. We then have the attack and release parameters of the compressor, which act as envelopes to shape the ducked reverb. We are affecting the reverb here and not the dry vocal line being fed into it. In essence, we are reshaping the reverb's response to adhere to the words triggering it. In essence, we're creating an envelope follower to alter the fixed response of the reverb. This allows us huge flexibility and control. So how do we do all this? The first step is to use a send and return or auxiliary for the reverb. If you insert the reverb on the vocal channel, then the compressor will also affect the vocals as well as the reverb and anything else in its path. Next, structure a reverb using simple parameters to create the start and decay of the reverb. Do this as you normally would for a vocal line. Place the compressor after the reverb and activate the external side chain. Feed the vocal line to the side chain of the compressor, so the vocal line will now trigger the compressor. Remember that you're not compressing before the reverb, you're compressing after the reverb structure, the compressor. I will show you the settings I've shaped for each parameter to help you on your way. But first, let's listen to the vocal line being fed to the reverb and then with the side chain triggered compressor. I'm gonna run through two examples using the same compressor settings but with two different reverbs, so you can gauge how effective this method is for both traditional vocal reverbs and the more esoteric tail based reverb delights. The two reverbs I'm using are Eventides 2016 and Unfiltered Audio's TAILS. First the dry vocal. And now with the Eventide 2016 without compression. And now with the compressor. You can hear how much clearer the start of the song words are with the compressor ducking the reverb. And another example, but using an ethereal reverb effect provided by Unfiltered Audio's TAILS. First the reverb version. I won't go into the reverb settings I've dialed in for both reverbs as there are only 365 days in a year but I will explain how to structure the compressor once you've created an auxiliary track. Place the reverb in the first insert slot, followed by the compressor. Make sure to use 100% send to the compressor sidechain, adjust the send amount to the reverb to taste. I'm using the following settings for the compressor and I've kept it nice and simple by using FabFilters Pro-C compressor. For style, I'm using clean. I'm doing this because I want a simple ducking response from the compressor. If you want to experiment, try using pumping as that will swell the reverb and give a denser response. Adjust threshold to taste and in this instance I've set it to -32dB. I've kept attack at the fastest possible setting, 0.005ms, but you can adjust this to suit the response you are after. Be aware of the different compressor topologies and how they behave as this will have a marked effect on the compressor's response. Adjust release to taste and I've timed it to be most affected at 17ms for this particular vocal delivery. I'm not using the parallel feature as I'm after simple ducking and nothing else. You can hear what a difference the compressor has made to all the above examples. The clarity of the initial vocal transients is unaffected and the reverb introduces itself far more smoothly than a simple triggered pre-delay. This technique works for any sound. I use it when applying reverb to guitars, pianos, anything plucked, percussive sounds and so on, there are literally no rules governing what you can use ducked reverbs for. To end this example, it is interesting to note that more and more manufacturers are including ducking as a feature on reverb plug-ins. In fact, Unfiltered Audio TAILS, which I've used here, comes with this very own ducker, but for obvious reasons I opted to use a compressor. In example four, I'm going to show you a very cool technique that we poor engineers use to make a lead vocal cut through a mix. The problem with optimising vocals within a mix is that you have all manner of shared frequencies to contend with. Vocals tend to encompass quite a significant frequency range, irrespective of gender and predominantly in the mid range, which is where most sounds dominate. In fact, people think the low end is difficult to manage. Quite the opposite. It's the mid range that is always hard to negotiate as so many sounds and instruments share frequencies within that range. Separating sounds and managing summing in this range is always a challenge because, you know, we don't have enough to deal with. It can be quite easy to end up in that abyss of tweaking whereby you raise the gain of the vocals and now it's too loud, drop the gain and it's being masked by other sounds. It's the same scenario with EQ. Boost the specific vocal frequency range and you're now altering both the timber of the vocals and the dynamics, which in itself can result in frequencies poking above other shared frequencies or masked by the same shared frequencies. So what do you do? Well, a very potent technique is to drop the gain of the mix with the vocal share frequencies. Only when the vocals are active and when the vocals are inactive, the gain rises back to where it was prior to the attenuation. We achieve this by using the side chain feature of a multi-band compressor, or in this instance, a dynamic equaliser. It's not very helpful ducking an entire mix to accommodate the space for the vocals to explore. It's far better to isolate the mixes frequency range where the vocals sit and duck that specific range rather than the whole mixes frequency spectrum. When using a vocal line to duck a whole mix, you need to plan the process sensibly. Simply ducking the mix at the master output will not work as the vocal line is also routed to the master bus and will also be ducked, so it's self-defeating. The trick is to create a sub-master and route everything to the sub master bar the vocals, which are directly routed to the master bus. The processed sub-master is also routed to the master bus. This configuration allows the sub-master processing and vocal line to output to the master bus, but separately. But before I dig into the settings, let's have a listen to the before and after examples. First, the dry version. And now with a dynamic equaliser. And with a multi-band compressor. I've processed this vocal line with FabFilter's Q3, which is a dynamic equaliser, and FabFilters MBC, which is a multi-band compressor, just to give you two ways of achieving similar but sonically different results. Once the sub-master has been created and everything bar the vocals routed to it, you can start to experiment with the ducking process. I started by inserting FabFilter's Q3 at the sub-master and activating its external sidechain. I selected the lead vocal line as the sidechain trigger and then honed in on the frequency range to duck. Normally you'd go straight for the range that is shared by the entire vocal line, but I chose to duck the low end of the synth sounds, top end of the bass and some of the hefty snare layers and to effect this, I chose a frequency range of 500Hz-3kHz. This is quite a wide range but it doesn't interfere with the frequencies I want to keep highlighted as this mix is predominantly low end driven. A Q value of 1.3 ensures a tight and focused bandwidth, but not too tight that you start to hear a notch effect as opposed to gentle attenuation. Use your ears to judge the bandwidth you want ducked. If you start to lose sounds, then back off and tighten the Q. FabFilters Q3 doesn't work off traditional ratios, but it does afford a dynamic range parameter, which acts similar to a downward compressed ratio. I went for a hefty -9.2dB range. This might seem over-excessive for usual mixed frequency game management tasks, but for this example it's worked really well. I use the auto threshold function as it's worked perfectly for this particular example, but please feel free to disable auto and input your own value. Q3 doesn't offer attack or release functions, but it is fast enough not to result in a lag, dragging response and not too fast to be choppy. Use the slope gradient parameter to instigate different compression / expansion responses. The result is a more controlled and robust vocal line that pokes above the mix without squashing the life out of the mix. Finally, I use FabFilters MBC to duck a similar range to the Q3's. FabFilters MBC is a multi-band compressor on steroids. It not only affords frequency specific compression, but also boasts a powerful expander, mid and side processing and various phase modes. I've selected dynamic phase mode which allows for a more transparent response, which is perfect for processing entire mixes. I've maxed the oversampling to four, which improves high frequency phase responses and helps to manage any aliasing that might creep in. Look ahead allows the compressor or expander to start reacting up to 20ms before game change is actually detected. This is an excellent way to preserve transients and avoid any aliasing that might occur when using ultrafast attack times. When I'm working with mixes or when I'm mastering, look ahead is almost always on. In this example I've dialed in slightly more pronounced values, notably the threshold which is set to -41dB. I did this to also duck some of the quieter sounds. I have made sure not to duck too heavily by selecting a 4:1 ratio and a small dynamic range of -6dB. Attack is set to 10% and releases at 20%. This allows attack transient to be left alone by the compressor and the body to be clamped down quickly and released fast, to allow the mix to jump back where it was after each ducking event. Adjust this to taste. The result is very similar to the Q3 ducking example, but it does sound ever so slightly different. You decide which you prefer. Well that's it for now, thanks for listening. This has been Eddie Bazil for Sound On Sound. Thank-you for listening and be sure to check out the show notes page for this episode where you'll find further information along with web links and details of all the other episodes. Oh, and just before you go, let me point you to the soundonsound.com/podcast website page where you can explore what's playing on our other channels.