It's a tall order this month, as we're asked to remix a reader's recording to sound like a big-budget Bond theme!
When I first heard Richard 'Wilx' Wilkinson's 'Dressed To Kill', the heavy influence of recent Bond-film theme-songs was immediately evident, but his original mix wasn't really doing justice to the music. Wilx had been commissioned to write the song for the closing credits of a tongue-in-cheek independent film about a hit-woman: hence the nod to 007. He'd created drum parts using Toontrack's EZ Drummer, adding synths, guitars, strings and a piano part (using Sampletekk's White Grand) to fill out the arrangement. Extra guitar parts were recorded by his friend Tom Warner, to give a stronger sound in the choruses. The electric bass in the choruses came courtesy of the EXS24 sampler — he'd have liked to use a real bass player, but simply ran out of time. Similarly, Wilx's vocal part was recorded quickly, due to time constraints.
The main problem for this mix was that Wilx's reference points were the single 'Butterflies & Hurricanes' (by Muse) and two Bond-theme singles: Garbage's 'The World Is Not Enough', and Chris Cornell's 'You Know My Name'. Not only are these all top-level productions, but they're also mixed and mastered phenomenally loud — so it's little surprise that, despite making significant progress with his mixing skills during the mix, Wilx felt there was, in his own words, 'an enormous chasm of difference'.
Wilx sent over the muiltitracks from his home in Brighton, and I loaded the unprocessed files into Cubase SX 2 and had a listen. A rough balance showed the verses coming across significantly more powerfully than the choruses — a problem that tends to be exacerbated when you drive a mix very loud during mastering, as severe limiting tends to push down more fully instrumented sections of the mix in favour of sparser sections. So my first job in sorting out the dynamics of the mix would be to try to get the chorus sounding as big as possible — after which I figured I could then find a balance for the other sections that didn't make the choruses sound small. Listening to the references, I felt that the Chris Cornell track would be the toughest to compete with. It was clear that I'd need to extract the maximum illusion of size from each of the parts, while using the minimum signal level, to allow the final master to be pushed as loud as possible.
Looking at the drums first, I wanted to be able to mix them comparatively low while still suggesting a powerful performance — which meant using compression, distortion and reverb to thicken the tails of the sounds, while keeping the low end and the signal peaks carefully controlled. The EZ Drummer tracks needed very little work to get a good basic sound, and in terms of equalisation, my main activity was high-pass-filtering all the tracks (kick, snare top and bottom, hi-hat, two tom mics, overheads, and room mics), to remove as much low end as I felt I could get away with (the low end of drum sounds really eats up headroom — it's fine for 'real' rock, but not what you're after for a more 'pop-rock' sound). I also low-pass-filtered the kick at around 2.9kHz, because the beater click was a little too prominent, and notched out a few unwanted lower mid-range resonances in the snare drum, overhead and room mics using very narrow peaking-filter cuts. The snare close mic still rang on a little too long for my liking, so I also expanded it a touch with Digital Fishphones' Floorfish, to shorten the tail.
Given the commercial ethos of the reference tracks, I decided to limit both the kick and snare tracks to keep them as even as possible. Many engineers use triggered samples to the same effect, but the EZ Drummer sounds were very tonally consistent as they were, and a soft-knee 250ms-release setting from Mda's Limiter plug-in levelled out the performance. I also used faster-release, hard-knee instances of this plug-in on the snare and hi-hat, to reduce the level of their initial transients — no drum transients were likely to get past the anticipated mastering here, so it made sense to focus on the body of each sound more than its attack.
For the overheads and room mics I wanted an aggressively compressed sound with lots of rock character. A Fairchild compressor emulation (the Antress Modern VFME plug-in) got the overheads pulsing a bit, while SSL's Listen Mic Compressor got the room mics leaping about in an angry kind of way. Bussing all the drum tracks to a group channel, and then further compressing this with Digital Fishphones' Blockfish (in its complex optical mode) brought the whole sound together into a single seething mass, with the tails of the kick and snare rushing up belligerently between hits.
A common rock mixing trick is to feed some of the drum tracks to distortion units set up as send effects. Units such as the Tech 21 Sansamp and Proco Rat often turn up in this role. I decided to give this a shot too, to sneak the drums' power and aggression up another notch. Two different sends were required for this, both of which used identical settings in Mda's simple Combo plug-in to supply the distortion. The kick and snare close mics fed one of these sends in mono, while the overheads fed another at a lower level and in stereo. These were high-pass filtered and mixed in to taste, and really helped to give more attitude, significantly increasing the contrast between the choruses and the verses.
A bit of short, bright chamber reverb on the kick, snare, overheads and room mics glued the overall kit together and increased the density and stereo interest — though a bit of high-pass filtering and a couple of EQ notches were needed in the return channel to avoid cluttering the mix.
The only other effects I added to the drums were to get the hi-hat and cymbals to sustain a bit longer, for which I used an eighth-note, tempo-sync'ed delay, limited and lightly saturated by Silverspike's Rubytube plug-in and fed on to the Room Machine 844 room-ambience simulator. With a bit of this effect added to the overheads and the hi-hat mics, I EQ'd the effect return to achieve the desired sustain timbre, in this case using a high-pass filter at 550Hz and cutting 5dB with a high shelf at 8kHz. This effect naturally changed the ambience for the other drums too, so I rebalanced the level of the chamber reverb in response.
In 'You Know My Name', I'd noticed that the guitars were usually mixed comparatively low (to make space for the vocals), relying more on the bass to fill out the texture, so I decided I'd try to do something similar, making the bass more prominent.
Whacking on lots of low EQ boost wasn't going to be much use here (because I needed to keep the low end of the mix very tightly controlled), so I tried to generate much more mid-frequency thickness by smacking the bass hard into an instance of the Antress Modern Analoger plug-in, which produced a much fuzzier sound — not that you heard it as a fuzz bass with the rest of the track going — and warmed up the low mid-range a lot, despite a 3dB low-shelving cut at 100Hz.
Another high-pass-filtered distortion send effect and a hefty 6dB EQ peak at 1.4kHz provided a bit more sonic edge, after which the sound was spread across the stereo image using Cubase's Symphonic modulation send effect. Modulation effects can really mess with a bass part's low-end stability, so I played safe by high-pass filtering the effect return.
The chorus guitars presented a challenge, because they had too much bite, which meant the notes of the chords didn't come through strongly enough — fine for punk, perhaps, but not here. EQ didn't help much, so I tried to add body using the bass guitar's distortion send and a fairly dull single-repeat tempo-sync'ed delay. This helped a bit, but more was needed, particularly given Wilx's referencing of the Muse track, so I decided to go slightly outside the Mix Rescue brief and record an additional guitar layer, double-tracking the chorus part to make the sound more 'epic'.
This left me with eight separate guitar tracks to mix during the choruses, and I immediately panned them across the stereo field to give a kind of 'wall of guitars' sound. It was then a fairly straightforward job to achieve a suitable balance using mainly high-pass filtering on the individual tracks, although I later moved the original pair of guitar parts further out of the way of the vocals, with a 3dB peaking cut at around 3.5kHz.
After graduating from a music degree in Liverpool, Richard landed a job working for Sony on the Singstar series of games, which gives him enough flexibility to work on his own projects: creating tracks for other people and backing tracks for musicals, producing and remixing bands and songwriters. He's currently composing for indie films in Brighton and London, and is getting into film sound recording and sound design while he puts his next showreel together.
The vocals had been recorded with an AKG C1000. This isn't my favourite mic for vocals, but it didn't sound too bad here; Wilx had probably added a big dose of high-frequency shelving boost on his mixer while recording. A little high-pass filtering was all I really needed to get the tonality in the right ball park, but I also wanted to create a very forward and 'hissing in your ear' sound for this song, so I reached for some heavy compression as well.
There are few compressors like Urei's super-fast 1176 for achieving this kind of effect, and following good experiences with the Antress LA2A emulation while remixing Imprint's 'Let Me Go' (SOS January 2008), I decided to try Antress' 1176 emulation, Modern VSME — and wasn't disappointed. I used a 20:1 limiting ratio with fast attack and release, adjusting the input gain control to pile on the compression, reaching more than 12dB gain reduction on some of the loudest notes. Sibilance is usually a problem when compressing heavily like this (and especially with a compressor like the 1176, which subtly distorts when working hard), but Digital Fishphones' Spitfish de-esser took the edge off this efficiently.
Similar EQ, compressor and de-esser settings were used on the four backing vocal tracks, but using the slightly less processor-intensive Cubase Dynamics and De-esser plug-ins. The latter tends to dull the overall sound a little (which is why I usually use Spitfish for lead vocals), but here the more understated high frequencies helped the backing vocals sit a little behind the lead.
There was enough going on in the mix already without vocal reverb, so I used a combination of a simple tempo-sync'ed quarter-note delay and a short 75ms slapback to sit the vocals into the mix. The former was EQ'd to favour the 1-4kHz zone and the latter concentrated a little lower in the frequency spectrum. For the slapback, I increased the delay modulation to maximum and sent the effect on through some compression to Cubase's fun little Grungelizer plug-in, which lent it the flavour of a rabble-rousing megaphone!
A bit of my standard Harmonizer-style stereo-widening effect (two different, short, pitch-shifted delays panned left and right) worked well here too, with slightly more making its way onto the backing vocals than the lead. I also dialled up one more distortion send effect for the vocals, carving away at its return channel with EQ to isolate some extra shoutiness. While working on the vocal sound, I took the opportunity to edit Wilx's vocal performances as well, to tighten up the timing between the parts, which seemed to me to be too loose for the style. Once I was happy with the timing of the lead, it was pretty easy to line up the other backing-vocal parts by eye — the general drum-and-guitars melée made even mid-syllable edit points very easy to disguise with simple crossfades.
The final piece in the chorus puzzle was the string section, which had helpfully been bounced down from Synful Orchestra in separate parts, so I had discrete tracks for first violins, second violins, violas, and cellos. I decided to go for a fairly dry and assertive pop-style string sound, aiming for the gap in the frequency spectrum above the guitar parts. I high-pass filtered fairly astringently, and compressed each part slightly, with fast attack and release times, to level out any rogue peaks. A dab of the delay/reverb patch I'd set up to increase the cymbal sustain helped the violin parts, in particular, to blend. But when I later sent a rough mix to Wilx, he said he'd had a more mock-classical sound in mind, so I lowered the high-pass filters a notch or two for a warmer tone and added in an extra reverb patch specifically for the strings using the SIR convolution plug-in. A 'wide ambient chamber' impulse with a pre-delay of 15ms gave a suitable sound (once I'd high-pass filtered it and dipped a little mid-range muddiness). The reverb time was a little on the short side, but SIR has built-in sample-rate conversion, so you can stretch impulse responses to achieve longer reverb times, and that saved the day.
A final stage in highlighting the rock element of the sound was to get some bus compression going over the whole mix. I tried various compressors (they all sound quite different in this role) and settled on another of the Antress plug-ins, the Manley Vari Mu emulation, Modern VRME. A medium attack time kept the bass response of the kick drum pretty much intact, while a fairly fast release produced a little beneficial pumping when the gain reduction was registering at around 4dB.
Having taken the choruses as far as I could for the moment, I turned my attention to the remainder of the mix, and because the lead vocal took on a greater importance during the verses I started there. The shoutier chorus sound wasn't really suitable to the more whispery delivery of the verses, so I immediately turned down the distortion send and switched off the slapback delay. The quarter-note delay needed to come down in level too, as the sparser verse instrumentation left it more exposed — though I did ride the level up occasionally to catch the odd word at the ends of lines. Where most of the vocal send effects came down in level, the stereo widener, conversely, worked really well at a higher level, diffusing the breathy menace of the verse vocal right across the stereo image.
One technical problem was that there were a lot of lip noises in the vocal recording (a common side-effect of boosting vocal EQ at extreme high frequencies) which were particularly distracting in the verses, so I spent a little time editing as many of these out as I could. I also tidied up a few areas of timing that were messing with the track's groove now that the vocal was taking up more space in the mix.
My next port of call was the layered pair of drum loops that provided the rhythmic backbone for the verses, though there were some EZ Drummer parts here as well. The loops had originally appeared under some of the chorus sections too, but I'd edited these out because they were taking away from the directness of the 'live' drums, and there was a similar problem during the verses, but in reverse — the live drums bogged down the rhythm, which reduced my ability to build the energy from the verses into the choruses. The main offender was the EZ Drummer kick drum, so I muted the close mic during the verses, and I similarly dropped the snare mic from the second verse to keep the live drums in a supporting, background role.
Out of the box, one of the loops sounded pretty good, giving a tight, compact sound, but the other was quite fluffy-sounding, ambient and ill-defined. I decided to keep this low in the mix, using some compression to get its peaks out of the way of the first loop while adjusting the tonality with some high-pass filtering and a hefty peaking boost of 8.5dB at 1.7kHz. A little of the kick/snare distortion send on each of the loops tied them together, making them harder-sounding and slightly crusty in a way that seemed to complement the vocal sound.
Listening more critically, I could hear a couple of overtones poking out of the mix unduly, and quickly sorted this out with narrow EQ notches. More of a problem was that the low end of the combined loop was still quite ambient, which not only undermined the mood but also reduced the potential for contrast with the roomier sound in the chorus. My solution was to roll off the low end of both loops and recreate the kick part on a separate track using copies of one of the EZ Drummer close-mic kick-drum hits. By limiting and filtering this new part fairly heavily, I was able to get a more suitable, tighter sound, plus the separate channel meant I also had the flexibility to balance this to taste against the rest of the loop.
The only other drum part in the song was a gong sample, which I pitch-shifted a few semitones downwards because it sounded rather 'half-fat'. A dollop of the cymbal's delay lengthened the sustain — and I liked the result so much that I copied the sample to the middle section, to imbue it with a subtle extra sense of drama!
During the verses the electric bass part was replaced with a bass synth. The raw file was quite dull-sounding, with lots of sub-bass, whereas the track needed much more audible mid-range energy to cut through. Adding mid-range EQ boost didn't do the job because some notes were less muffled than others (probably a by-product of the velocity response of the synth patch in question). Loading up some stiff processing from Cubase's Multi-band Compressor helped here, and I split the sound at 80Hz and 250Hz to balance the different lower-frequency regions better. Even after this I still fancied a bit more sustain, so followed up with full-band compression from the Dynamics plug-in, being careful to set the time constants long enough to avoid unmusical distortion artifacts (which can be a bit of a problem with this particular compressor).
This kept the synth firmly nailed to its location in the mix, and made it much easier to hear on smaller speakers. As a final touch, I sneaked on some of the Symphonic effect I'd already set up for the chorus electric bass sound and supported the sound's built-in delay effect with some of the quarter-note vocal delay.
Around the bedrock of vocals, bass and drums, the remaining guitar, synth and piano parts all fitted in very easily, with only high-pass filtering, the odd EQ notch, a send or two to the delay effects and some level automation. There were a couple of exceptions to this, but still nothing complicated: two separate guitar parts were playing in the verses, so I treated one of them with a phaser to make it more distinct from the other, and I also slightly overdrove the reverse guitar effect that precedes the choruses, to give it more urgency.
I'd been comparing the verses and middle section with the choruses as I went, and although I was pretty happy with the sound of each section independently, the overall energy levels weren't quite flowing right from one to the next, which robbed the choruses of impact. Careful automation of the drum, bass, and vocal levels helped a great deal here, and I also inserted GVST's little GGain plug-in into the master output before the buss compressor, gently riding its level to drive the compression harder during the choruses.
I also made a few little arrangement tweaks to keep the interest going in a couple of spots. For example, a reversed snare hit just after the first chorus made the entry of the second verse more dramatic, while the little distorted vocal delay effect at the end of the middle section helped keep the momentum going into the final choruses. The string phrase in the introduction seemed a little overbearing coming from the whole string section, so another thing I did was cut out everything but the violas and send from that track to a few of the send effects, to produce a shiftier and more mysterious feel.
Wilx: "While recording 'Dressed to Kill' I made the decision to use it as an exercise in improving my mixing skills, so I pimped a few rough versions around various forums to get some feedback and advice. When Mike suggested I send the track in for Mix Rescue, I was eager to hear what my song would sound like after being disassembled and then stuck back together by someone with more mixing experience.
"I'm delighted with the result! My main concerns about the track had been that it lacked 'balls', that the dense arrangement sounded too cluttered, and that it didn't sound like a cohesive whole. All of these points have been significantly improved. My choices of comparison material may have been somewhat over-ambitious, but I think Mike has done a fantastic job, taking the mix to a much more polished-sounding level than | thought would be possible.
"The song has more clarity and space. Everything 'sits' in the mix more comfortably, and the fullness of the production — both dynamically and spatially — is much better.
"I've still got a way to go before I'll be perfectly happy with my mixing abilities, but something like this remix helps immensely, and I hope the rest of the SOS readership can pick something up from the article too. So thanks Mike, and thanks Sound On Sound!"
After tweaking the night away I'd arrived at a mix I was pretty happy with on its own terms, but there was still work to do in terms of matching the overall tonality and volume against Wilx's chosen reference tracks. For this task I bounced down my mix as a 24-bit WAV file and reimported it into a new Cubase Project alongside the reference tracks.
Even with my mixdown normalised, I had to turn down the reference tracks by 7-9dB before their subjective loudness matched even the hottest section of my raw mixdown file: someone had clearly been setting the loudness maximisation to 'stun'! Also, the tonality of the reference tracks had clearly been hyped at the high and low ends, presumably to translate better at low volumes and on small speakers, but the amount of low sub-bass had been deliberately restricted to free up mastering headroom.
I'm no mastering engineer, and I don't have access to the specialist facilities that the challenging task of mastering requires, but referencing my mix against tracks like these would have been impossible without approximating something like a similar mastering chain. The extreme nature of the processing all too often used on commercial singles like these can have drastic effects on the mix balance, especially on a complex arrangement, as in 'Dressed To Kill', and I needed to know whether I might need to compensate for this with some adjustments to my mix.
Using Buzzroom's Genecomp 3 multi-band compressor, I went in search of the 'hype' by compressing in three bands with comparatively high ratios (high for mastering, that is!) of 1.3:1 in the central band and 1.4:1 in the outer bands, all in soft-knee mode. The idea behind the different ratios was simply to compensate a little for the drop-off in the ear's sensitivity to high and low frequencies at the kinds of lower listening levels for which radio singles tend to be destined.
Short attack and release times meant the average signal level was raised fairly strongly in each band, but also resulted in a middle-heavy sound (Genecomp 3 doesn't automatically make up the gain lost due to its compression). Boosting the low and high bands by a couple of decibels improved things, but I had to adjust the crossover points to get the best results: I'd normally choose something like 100Hz and 5.5kHz for more subtle polishing, but I had to move the lower crossover up to 240Hz before the low mid-range had the slightly tubby prominence I was hearing in the references. In tandem with this I high-pass filtered before the compressor with Cubase's Q plug-in, to roll off as much of the sub-bass as I felt I could get away with, pushing the low compressor band's gain up a fraction to compensate with more lower mid-range energy.
I'd already achieved a significant loudness increase but needed to push things further to get the sound of the reference tracks. Buzzroom's BuzMaxi brickwall limiter worked well when listening to a louder section of the mix, but the limiting action brought up the verse levels more than the choruses, destroying the mix dynamics. Mix dynamics inevitably suffer when you push things this loud (the reference tracks had similar problems), but there was no doubt it took things too far here.
My solution was first to set up a multi-band limiter patch using Cubase's Multiband Compressor plug-in, adjusting threshold levels to try to shave another decibel or two off peaks in five different frequency bands — the same kind of principle used in a number of dedicated multi-band mastering limiters. Before that, I also gently squashed the signal with GVST's GClip soft-clipper, trading off some subtle distortion side-effects in return for less limiting. These settings enabled me to back off the BuzMaxi settings from around 8dB of limiting to only 2.5dB, which left the mix dynamics in much better shape.
As I'd suspected would be the case, after I had applied the mastering processing a number of mix levels needed adjusting to maintain the right balance, so I set about tweaking and remastering the mix a few times to pull things into line. In particular, the kick and snare levels needed adjusting, as the limiting punched them into the mix, and the levels of the bass guitar and bass synth were being pulled up too high during the verses. The upper end of the cymbals was another offender, becoming a bit Kate Moss (rather fatiguing after all the hype...) Some compression of the overhead frequencies above 8kHz and two small EQ dips at 5kHz and 8kHz smoothed things out again without too much trouble.
While working on these nips and tucks, I noticed that the stereo images of the reference tracks (especially the Chris Cornell one) were significantly wider than that of my mix. I tried applying GVST's GStereo multi-band stereo manipulation processor as a mastering process, but this global processing only worked for some parts, so I applied only a very subtle setting globally and went back to the mix to tweak things on a track-by-track basis. In the case of some parts, such as the guitar, it was just a case of panning tracks wider than before, but with some of the stereo synth parts and effect returns I had to use GStereo instead to reduce the mono component of the stereo signal.
These changes had left the vocal and bass rather high and dry in comparison with the rest of the track, when I checked the mix in mono. This is why I'd kept my mix a bit narrower: summing panned stereo mixes to mono causes centrally panned sounds to increase in volume by roughly 3dB relative to hard-panned sounds, because of the way stereo speaker systems create phantom central images. However, this sound was still closer to the references, so that was the way I left it.
In the end I managed to get this mix to compete pretty well with the references Wilx provided, given the limits of the facilities I had available. More importantly, though, I felt that the mix would be more likely to stand up to fairly brutal mastering processing without falling apart if he chose to get the raw mix remastered at a later date — and in the interim my rough-and-ready maximised version would serve fine for his own demo purposes. One final note of caution if you're thinking of trying out heavy mastering processing at home, though: always remember to save an unprocessed version of your final mix as well as the processed one. That way, no matter how expensive your processors or how good you think it sounds, you always have the option to remaster at a later date — something that could be particularly important if you need to make it sit with other mixes on an album or compilation.
Thanks to guitarist Tom Adams for the guitar overdubs, and to Hills Road Sixth Form College, Cambridge, for the kind loan of their studio facilities.
It's a lot easier to appreciate the techniques Mike describes in this Mix Rescue if you listen to the 'before and after' audio examples we've placed on the SOS web site at www.soundonsound.com/sos/feb08/articles/mixrescueaudio.htm.