One element of the vocal arrangement that resisted this approach was the chordal backing-vocal 'oohs' in the choruses. Things like reverbs and delays don't really help unmask those kinds of sustained sounds, so the voices were getting lost amongst the other sustained synth and string parts. Fortunately, there's an easy way of drawing more attention to the individual voices in any chordal pad: stagger their timing. If all the notes in all your sustained parts start and/or move on the downbeat of the bar, those changes won't stand out nearly as much for the listener, but if you introduce some of those notes on different beats, then they draw the ear better. So in this scenario I just edited Roger's backing-vocal lines to create a slow broken-chord figuration, and also held back some of the higher lines for later during the chorus section to improve the sense of build-up.
Another concern was that the vocal arrangement appeared to lose momentum during the gaps between lead vocal phrases, despite the lead part's trailing rhythmic echoes. Fortunately, there was a nice little scalar backing-vocal moment in the second verse that I felt deserved more exposure, so I experimented with copying that to fill in those chorus-section lulls — it worked a treat, especially with a little quarter-note echo to turn the original falling scale into a cascade of falling thirds. In fact, I liked the falling harmonies so much that I plugged the line into Celemony Melodyne to generate upper and lower harmonisations too, and mixed those in to intensify the effect towards the end of each chorus. While I had Melodyne open, I also pitch-shifted one of the lead-vocal melismas and used it to add further higher-register ululations to the chordal backing-vocal arrangement.
Pad sounds can be a mixed blessing. On the one hand, they can make pretty much any arrangement sound fuller and warmer; but on the other, their comparatively static nature can leave a mix lacking organic detail and musical momentum. In Roger's case, he'd filled up the chorus texture with three main 'pad' elements: the upper spectrum of the main synth-bass part; a synth strings patch; and the backing vocal 'oohs'. Although the new vocal arrangement had drawn more musicality from the last of these, the first two still felt like they were blanketing the mix and undesirably homogenising the overall sonics. I decided to tackle this in a couple of ways.
Firstly, I used some rhythmic effects to increase the sense of modulation. For the bass-synth element, I used TAL's cool little freeware Filter II plug-in, drawing in a quarter-note ramped modulation curve for its 24dB/octave low-pass filter, and then adjusting the filter resonance and modulation depth to taste. My aim there was to try to get some useful movement and rhythmic support, but without anyone thinking I'd got a new autowah plug-in for Christmas! For the strings, I used audio edits in conjunction with long fade-ins to duck the pad on the downbeat whenever it changed chord.
My second tactic was to supplement the existing pads with a variety of subtle sound-design tracks, in order to make the pad texture more organic and involving over time. The main trick when doing this is not to think in terms of just adding further continuous 'layers', because it's moment-to-moment sonic variations that you're trying to add. One approach might be to add one or two new parts where you've gone crazy with automation and modulation data to continually mutate their contributions throughout the timeline. Personally, though, I find I get quicker results by taking a kind of jackdaw approach based on multiple synth presets. Let me explain what I mean...
What I do is simply surf through the presets of any likely-looking virtual instrument in my DAW (I used mostly Heavyocity's Evolve and the bundled sound library in NI's Kontakt), searching for patches that have some sense of 'movement' to them, whether that's sequenced rhythmic elements, swells, arpeggiations, modulated parameters, or any other type of time-varying characteristic. When I find a preset that sounds appealing, I use it to record the most obvious ideas that come to mind, based on the nature of the preset and how that fits within the mix context. I don't try to cover the whole chorus section, but just drop it into the one or two spots where it most seems to make sense. So if there's some kind of swell baked into the synth patch, then I might just play a note or two to make that synth swell towards an important beat or bar line. Or if I've chosen an arpeggiated preset, I might have that bubble up in a gap between vocal phrases, or echo a cymbal hit. Or if I've selected some kind of sequenced loop patch, I might record that for a few seconds, bounce it as audio, and then fade it in and out to add some stereo rhythmic interest between the other sound-design elements.
The point is that I don't think too consciously about what I'm creating at the outset — I just repeat this process with a bunch of different presets (bouncing down the results to audio as I go so I don't overload my CPU), effectively throwing a load of mud at the wall to see what sticks. However, what I find is that the more of these sound-design elements I add, the more it begins to become apparent which ones are the most musically appropriate at any given moment, and which ones I should prune away to let more interesting things through. And once I've finished that editing process, the result is a 'patchwork quilt' of different synth atmospherics that's tailored to the flow of the music in an organic way, because it's based on my own human reactions to what I'm hearing.
In this specific case, I wove together about 15 tracks of these little synth embellishments for the song's chorus section, all of them fading in and out of the texture at different moments to add the kind of patina and complexity I was looking for in the synth backing. That might sound like a lot of tracks, but all of them were very low in level and functioned less like extra musical parts and more like a subtle send effect, in that you only really noticed their presence if you muted them! Creating so many additional synth parts might sound like an awful lot of work, but it's not nearly as onerous as it seems, as you're not actually doing any fancy sound design; they're mostly one-finger keyboard performances, and the 'patchworking' process involves nothing more complicated than audio fades.
Additional virtual instrument layers turned out to be useful elsewhere too. I noticed, for instance, that the hi-hat part wasn't adding much in the way of rhythmic detail to the drum programming, and there were no percussion parts amongst the multitracks either. So I decided to layer in an abstract rhythm loop from Ueberschall's veteran Groove Shadows library to add some lower-level detail to the drum patterns, a move that also seemed to make the kit as a whole sound a little more cohesive. During the third verse I supplemented Roger's string parts with some more expensive-sounding VSL contrabass and viola lines, and also expanded on the handful of ear-candy elements he'd added using further synth effects, short harp arpeggios (with accompanying reverse-delay), cymbal/gong rolls, and finger cymbals.
As is often the case in mixing, it wasn't actually simple 'mixing' considerations that were holding this project back. Yes, the low end and the vocal sound both presented processing challenges, but arrangement and sound-design work played an equally important role in enhancing this project's timbre, detail and musicality.
So the next time you feel that your processing plug-ins are letting you down, perhaps consider putting them away and breaking out the audio-editing tools or an extra synth instead.
This month's song is 'Muddy Water', by Norwegian singer-songwriter Dorothea Wessel. The acoustic guitar part was played by Emily Zimmer, but all other recording and programming was carried out by Roger Wessel.
Most modern DAWs offer sophisticated facilities for time-stretching and tempo-matching, via a system where 'warp markers' in the audio can be dragged to manipulate the timing of specific audio events; the audio between the warp markers is automatically time-stretched/compressed as required. As quick and convenient as this method is to work with, it makes heavy demands on the underlying time-stretch algorithms, and it frequently leads to unwanted sonic side-effects such as loss of transient definition and/or a sporadically chorusey tonality. This is precisely what Roger encountered while tightening up the timing of the main verse acoustic guitar part.
So for this remix I took a different corrective approach — I resorted to standard audio-editing tools to move slices of audio around as required. Once I was happy with the timing, I crossfaded any resulting gaps or overlaps in the audio, occasionally tweaking the crossfade shape or the phase relationship between the incoming and outgoing waveforms if any of the edits sounded 'lumpy'. The result was no less 'in time' than Roger's, but it avoided any time-stretching side-effects, which meant it sounded much clearer and cleaner, even before I did any mix processing.
That's not to say that my editing approach had no side-effects at all, though — I wasn't able to disguise a couple of crossfades as well as I'd have liked. Within the context of the full mix, though, I don't imagine anyone will notice those edits but me, and in the grand scheme of things they're a small price to pay for a dramatic improvement in the overall guitar timbre.
As part of this remix, I used quite a lot of bus compression to help create a sense of balance interaction within the ensemble — something that doesn't occur naturally when your parts are all programmed and overdubbed, as in this case.
The drums in particular went through a Waves API 2500 on their own group bus, as well as a second API 2500 and an instance of Cytomic's The Glue on the master bus. While I was happy with the character and movement that these compressors added to the mix, I found myself losing more transient and punch than I'd have liked, so I resorted to a little dodge I often use under such circumstances: I mixed some of the uncompressed kick and snare sound direct into the master outputs, bypassing all three bus compressors.
This ploy won't work properly, though, if the kick and snare signals that pass through the main mix processing incur any significant phase shift — and that fact that very nearly bit me in the arse on this remix! You see, I set up my direct-to-output sends for the kick and snare midway through the mixing process, checking them for phase-alignment at that point.
Later, at the mix-referencing stage, I adjusted my master-bus processing, without consciously directing my attention towards how that affected the phase-match of the direct-to-output kick and snare. It was only at the last moment, in fact, as I was going back through my settings in preparation for writing this article, that I realised what I'd done. Luckily, I preferred what the phase shift had done for the kick timbre! But the direct-to-output snare was definitely hollowing out that drum's lower mid-range undesirably. My fix was to apply a simple 110Hz high-pass filter into the direct-to-output snare feed, which actually gave a much fuller snare sound, counter-intuitively, by virtue of the filter's phase-shifting side-effects.
Roger & Dorothea Wessel: "This mix turned out quite different to how we imagined — and we both love it! There's a real contrast between the spaciousness of the intro and the really close first verse, and the guitar on the verses sounds so much better than in the original mix. But what strikes us most is how you've managed to get the chorus so spacious and huge, and also how you've made the transitions from the closeness of the verse into this huge chorus in a beautiful way."
You can hear my remix of this project on the SOS website at sosm.ag/mix-rescue-0320, alongside a selection of audio examples demonstrating many of the mixing and arrangement techniques showcased in this article.
Or download the MP3+WAV files saved in this ZIP file: mixrescuemarch2020.zip
Here's my final remix of Dorothea Wessel's 'Muddy Water', carried out on my own Cockos Reaper-based DAW system in conjunction with third-party plug-ins from Audio Damage, Blue Cat Audio, Brainworx, Celemony, Creative Intent, Cytomic, Dead Duck, Fabfilter, GVST, Infected Mushroom, Izotope, Klanghelm, Lexicon, Martin Best, Melda, Native Instruments, TAL Audio, Tokyo Dawn Labs, Toneboosters, Schwa, Sknote, Slate Digital, Sonalksis, Sonimus, Stillwell Audio, Ursa DSP, Valhalla DSP, Voxengo, Waves, and Wavesfactory.
If you'd like to check out all my processing and effects settings in detail, then you can download the full Cockos Reaper project file for this month's remix at www.cambridge-mt.com/ms/mtk/#DorotheaWessel — and you'll also find the project's raw multitracks there too, if you fancy trying your hand at this mix from scratch!