Performed live and filmed in a single take, Justin Timberlake’s ‘Say Something’ is one of the most ambitious music videos ever made.
Just before Christmas last year, French sound engineers Henri d’Armancourt and Guillaume de la Villéon spent five days in Paris mixing a new single for Justin Timberlake. They received little feedback, apart from a request to come to New York — and when the French duo duly set up at Jungle City Studios, it turned out that Timberlake and his musical director, Adam Blackstone, had not even heard the rough mix. Luckily for the two Frenchmen, Timberlake and Blackstone reacted with an enthusiastic “Incredible!” followed by “And how did you do it?”
Watching the official music video for ‘Say Something’, the third single from Timberlake’s new album Man Of The Woods, is likely to elicit the same response — at least, once it dawns that the entire video is done in one shot, with the audio recorded live on set. D’Armancourt and de la Villéon acted as sound directors for the video, and later mixed the audio. If you haven’t seen and heard the music video yet, take a moment to go over to YouTube, and prepare some support for your jaw, as it’ll almost certainly drop.
The 6:22 video for ‘Say Something’, featuring Chris Stapleton, is by all standards a tour de force. It’s set in the Bradbury Building in Los Angeles, first made famous by its appearance in Blade Runner in 1982. Both ‘Say Something’ and the iconic Ridley Scott movie make ample use of the building’s labyrinthine corridors, stairs, balconies and elevators, as well as its huge atrium. The music video starts with Timberlake in a small room, pushing buttons on a Native Instruments Maschine controller to trigger samples from the studio version of the song. The real action begins at 1:13 when the sound of an acoustic guitar enters and Timberlake walks down a corridor, humming, passes the guitarist, enters an elevator, and picks up his own guitar.
Emerging from the elevator a few floors up, Timberlake passes a pianist before the camera pans to reveal Adam Blackstone playing bass on the balcony, and then settles on country singer Chris Stapleton, who is on the other side of the building, also singing and playing guitar. The choreography grows ever more complex, with Timberlake and Stapleton walking around, stepping into two elevators, and meeting up, still singing and playing guitar. The climactic moment sees a 54-strong choir join from the balconies. In the vast majority of music videos, there’s little relationship between the audio and what happens in the video; by contrast, the connection between sound and vision is total in the ‘Say Something’ video, giving the viewer an exciting sense of witnessing an event taking place in real time.
The creative forces behind the video belong to French production company La Blogothèque. Back in 2006, independent film-maker Vincent Moon and producer/writer Christophe ‘Chryde’ Abric sought to develop a new music video format they called ‘Concerts à Emporter’, or ‘Takeaway Shows’, which would be filmed and recorded in a single live take. Today, La Blogothèque employ eight people and uses a pool of film-makers and sound people to create their music videos, which now also include the Pocket Parties series, which uses multiple cameras, as well as large, major artist, concert recordings. The Takeaway and Pocket Parties series are posted on the company’s web site and YouTube channel.
It was the Takeaway Shows format that attracted Timberlake and his crew, and by the time d’Armancourt and de la Villéon were invited to join the team, film director Arturo Perez Jr and four people from La Blogothèque, including creative director Christophe Abric, were already in Los Angeles, working out how they were going to do the video. Their original idea involved recording four or five musicians in Ennis House, a 100-year old building designed and built to look like a Mayan temple, but by the time d’Armancourt and de la Villéon arrived in LA on Monday December 4th, the concept had grown far more ambitious.
D’Armancourt recalls: “After we arrived in Los Angeles, we went from surprise to surprise. The first thing we were told was that they had decided on the ‘stripped-down version’, with 17 musicians in a big five-storey building. We were like, ‘How’s that stripped-down?’ On Tuesday we met with Justin Timberlake’s team and director Art Perez at the Bradbury Building. We’d been told that Justin’s team might want to put up all the mics and we’d just do the ambient mics, but Art announced us as the sound directors, so that’s what we ended up being. On Wednesday we visited the Bradbury Building and met Justin Timberlake and Adam Blackstone, and all of us walked through the building. By the end Justin said, ‘Hey, what if we add a choir of 50 people?’”
De la Villéon: “On Thursday we called around rental companies in Los Angeles to get all the gear. We also were present at band rehearsals on Tuesday and Thursday. Things kept changing and a last-minute decision was to put the horns on the second floor. On Friday we set up the gear, and there was a general rehearsal without the choir. On Saturday we shot five final takes during the daytime and one during the night. The first two takes were not as good, the next two takes were pretty good, and the fifth one was great. They were very happy with that last daytime take, so they knew they had something good. Then we set up for the night version, and we only did one take of that. They preferred the night version, which was more mysterious and more magical, and it’s the take that was used.”
The challenge, then, was not only to record Timberlake and Stapleton singing and playing acoustic guitar while walking around and going up and down elevators in a huge building, spread out over five floors. In addition to the two main characters there was also an electric guitarist, a third acoustic guitarist, a percussionist, a drummer, a piano player and a Fender Rhodes player, four horn players, four backing vocalists and the 54-singer choir. D’Armancourt and de la Villéon stress that many of the solutions they found were based on know-how developed at La Blogothèque over the last decade by the in-house team of eight sound directors. Clearly, tons of wired and wireless gear would be required, and d’Armancourt and de la Villéon’s first request was for a radio frequency coordinator familiar with the US situation.
De la Villéon: “We did not only need to be able to record all these musicians across five floors, but also provide all of them with monitoring — plus we needed all the audio to be routed to a monitoring console, from which it was sent to the in-ear monitors, a broadcasting console for a live preview mix of each shot, and the audio recorders, which were two Sound Devices 970s, because they are exceptionally reliable and never crash. We ended up using an Axient digital wireless system for the transmitters, and an analogue wireless for in-ear monitoring. The heart of our recording setup was Audinate’s Dante, or Digital Audio Network Through Ethernet, which uses standard Ethernet RJ45 cables, and works with switches that allow you to plug in as many devices in as many directions as you want. On the different floors everything that was wired was plugged into Audinate’s stage boxes.”
D’Armancourt: “In total we had 64 recording channels that went to the Sound Devices 970s. We had 16 digital Axient wireless transmitters, so we had to choose wisely what to use them on. Of course, when you’re moving around you need wireless, so Justin and Chris each had wireless transmitters for their two DPA 4060 lavalier vocal mics and their guitar. For the rest, wireless transmitters were used for the percussionist, who had a DPA 4060 mic on each wrist; each of the four backing vocalists had a DPA 4080, and each of the two elevators had a Neumann KM184 ambient mic with wireless. Justin’s crew provided most of the band instrument mics and DIs, though Blogothèque took care of a drum DPA 4061 ‘underhead’, two piano DPA 4060 mics, acoustic guitar DPA 4060 mics, the vocal mics, and all the ambient mics.
“We in general use many DPA 4060 mics, because it’s the best-sounding lavalier mic out there. The only disadvantage is that it’s a bit larger and harder to hide. On Justin and Chris, the lavelier mics were stuck between two buttons of their shirts, and we had to pray that Chris’s beard and chest hair did not rustle, but it turned out fine. Just to be sure, we had a plan B that consisted of us attaching a Countryman B6 mic to his hat, pointing down at him.”
De la Villéon, meanwhile, explains what d’Armancourt and he were up to while shooting. “I was in charge of the Dante network. During the recordings I was at the broadcast mixing desk, doing live mixing, which was similar to what I do most of the time, which is front-of-house mixing. So when they wanted to look at takes, they were listening to my mix. I was also interacting with Justin’s monitor sound engineer Paul Klimson, who was in charge of the gain and sending signals to the in-ear monitors. Paul and I were working together to check all the time that all the sources were OK. For example, at one point there was a buzz in one of the wireless mics, and we also had a problem with a piano mic, which was a faulty cable. So these problems had to be addressed immediately. We also had a local sound engineer making sure all the Sound Devices recorders were working properly.”
D’Armancourt: “In addition to the 64 recording channels that came in via wireless or cabling, we also had another 10 recording channels that went to two mobile recorders, the Sound Devices 788 and the Sound Devices 633. The two boom mics were recorded on these, for example. There was a boom mic operator recording Justin as he walked from the Ableton setup to the elevator, but together with the Steadicam operator there was not enough space in that first elevator, so the boom operator had to stay outside, and all that remains in the elevator are the wireless mics. I was the second boom operator, and waited on the fifth floor for Justin to come out of the first elevator, and I then managed to follow the Steadicam until the end, also going into the second elevator. The camera and all Sound Devices recorders are sync’ed with Digital Clapperboard time code.”
The late addition of the choir provided d’Armancourt and de la Villéon with some challenges, as the former explains. “We didn’t get to rehearse with the choir, so we did not know what it was going to sound like in the building. But because the choir was so spread out, we could not have two or three mics in front of the singers, and we had to record them with the ambient mics. The choir director had in-ear monitors, while the choir itself could hear the snare drum for timing, and there was plenty of sound from the electric guitar and the horns for them to get pitch information. However, we were aware that the snare drum was a bit overwhelming in the ambient microphones, and so in case that would later give us problems, we did some takes just with the choir, as a backup. That was really helpful later during mixing.”
De la Villéon: “It’s always the same thing, which is that the human voice always loses in competition with the drums, electric guitar, and so on. So yes, you could say we cheated with the choir. But the problem with the snare sounding around the building really affected what we recorded with the ambient mics. Because of the distances in the building, there were ambient mics on which the snare sounded a full beat late! The snare really prevented us from using the ambient mics as much as we normally would. The snare also was in the vocal mics, because the DPA 4060 is omni. You can hear the snare in Chris’ voice when he starts singing.
"Our ambient mic setup consisted of two [Sennheiser] MKH8020 room mics, and five ORTF setups, which is a technique from Radio-TV-Française, using two cardioid mics that are 17cm apart, at an angle of 110 degrees. On two ORTF setups we used Sennheiser MKH8040s and the other three had Neumann KM184s. We also had four Sennheiser Ambeo four-capsule mics set up, but in the end these were less important.”
After the first night take, the decision was made to do no more takes. “Everything was perfect with that one night take,” recalls d’Armancourt. “The night version had this unexpected thing with the elevator, with the lights flickering to give a kind of Alfred Hitchcock effect. It was pretty staggering visually. The cinematography team was super creative and did a great job.”
De la Villéon adds: “Everyone left to party soon afterwards, but Henri and I remained because we were doing backups for hours. We made two copies of all the sessions, and in the end we had three times 300GB. One copy stayed in the US and we took the other two with us to Paris.”
D’Armancourt and de la Villéon conducted the mix of ‘Say Something’ over five days, December 18-22, at the former’s studio in Paris, called Velvet Underground Studio. “It’s a home studio,” says its owner. “I had put up the famous poster of the Velvet Underground with the banana there, and thought it was a good name, also because the studio is underground, and there are velvet drapes in the live room to treat it acoustically.”
“The studio is a bit more than a home studio,” interjects de la Villéon, “because it’s used professionally. It’s a big room and sounds extremely good. Henri has K&H O300 speakers, which I love, and when we later went to Jungle City Studios, it sounded very similar.”
D’Armancourt: “The first thing we did was to put all takes in one session, on different playlists, and check that everything was synchronised with the visuals. We also wanted to make sure that if they suddenly decided that they wanted the day version after all, that we could switch to that immediately, with us being able to keep our settings on each track, and plug-ins, and also the automation.
“Before we even started to mix there was the huge job of cleaning all the audio. We heavily relied on iZotope’s RX De-Noise Spectral Audio Repair and De-Click plug-ins, and we also used SPL Transient Designer and dynamic EQ plug-ins to prevent spill from doing weird stuff.”
De la Villéon: “The biggest challenge we faced was to retain the huge sense of space that was present in the Bradbury Building in the audio, without having a sound full of dirt, far away and not precise. We began our mix with working on the ambient microphone tracks, spending even more time cleaning up all the frequencies and all the dynamics. It was a massive job to be editing and trimming everything to only the bits of audio that we wanted to use. If we’d left all tracks open all the time, it would have sounded like a complete mess. We did this on all tracks. If there was a space between two words in the backing vocals, we’d take that out to prevent spill.
“Once we had cleaned up the close-mic tracks, we put them together with the ambient mics, chose which ambient mics worked best, and adjusted everything to the right amount of ambience, while also deciding whether we needed to add things like reverb.”
D’Armancourt: “Because of the delay and spill from the snare in particular, there were three things that we did. First of all, we used close mics as much as we could, especially on the leads. The second thing is that we put all the ambient mics in phase with the instrument that’s spilling the most into it — again, mostly the snare. This also involved a lot of fader riding, pulling the fader down the moment you don’t want to hear something any more. We did this far more than we would normally do in a live recording. We then delivered a preliminary mix, which used the techniques that we normally use in Blogothèque mixes, with a lot of ambience, a bit harsh, quite roomy, and you could really hear the building. After internal discussions with Arturo and Christophe we decided that it should sound more like a mainstream mix, more dry and focused, and we delivered a mix on December 22 in which we went for a sound that Timberlake’s fans would be able to relate to, while still remaining true to what you’re seeing in the video.”
The ‘Say Something’ Pro Tools mix session that d’Armancourt and de la Villéon worked with consists of 105 tracks. All the raw audio tracks are grouped and routed to aux tracks, with the latter usually appearing immediately underneath. So for example, after the video track at the top, there are four tracks with the Ableton material that plays at the beginning of the video, which in turn feed an Ableton aux track just below it.
Following the session further from top to bottom, next are two grouped pre-recorded percussion audio tracks, and the pre-recorded percussion aux track. Next are two pre-recorded music tracks, with background effects, with a music aux, two effects tracks (with the sound of glass), and an effects aux, two unused backing vocal tracks with an aux, and the Maschines aux, to which all the above-mentioned aux tracks are sent. Next down are a snare trigger track, a kick, a kick trigger, two real snare tracks and overheads, all going to a drums aux, followed by two percussion tracks and a percussion aux. A bass DI track follows, and then six guitar tracks, consisting of two electrics, the three acoustics played by Timberlake, Stapleton and Ives, and another acoustic guitar track, all sent to the guitar aux. Stereo piano and Rhodes have their own auxes, as do the four horn section tracks. Finally, a ‘Music All’ aux combines all the instrument tracks.
Tracks 49-69 are vocals, comprising backing vocals and associated aux, Timberlake and Stapleton lead vocals and a leads aux, and then a further global Vox aux track that receives all the lead and backing vocals, along with the six choir tracks, via their four related auxes.
The next section, 70-91, consists of all the ambient mic tracks with associated auxes: AB stands for two omnis, the ORTF tracks are numbered to indicate which floor each of them was positioned at, and there are various tracks from the Sennheiser Ambeo mics. The last section consists of effects aux tracks, mostly hosting reverbs such as SoundToys’ Little Plate, HOFA’s IQ-Reverb and Audio Ease’s Altiverb, plus an ‘FX All’ aux. Track 104 is the master, containing only two plug-ins, and the session is completed by the master print track at the bottom.
It is notable that large parts of the session contain almost no plug-ins. There’s nothing on the Ableton tracks, the effects tracks, the early backing vocal tracks or the Maschine track, while the pre-recorded percussion and pre-recorded music auxes have a single EQ each. However, some tracks have been more heavily processed, notably the two live snare tracks, the bass DI and guitar tracks and, most of all, Timberlake and Stapleton’s lead vocals and lead vocal aux, which have tons of plug-ins and sends.
De la Villéon: “We put the [FabFilter] Pro-Q2 on the kick to cut a frequency around 150Hz and help it cut through a bit by adding 5kHz. It also has the Waves CLA-76 [compressor] and Waves SSL E-channel [channel strip]. The first snare track has the Waves SSL G-channel EQ, again the CLA-76 and also the SoundToys Devil-Loc [compressor and distortion effect], because it was a side-stick and we wanted a little bit more crunch to help it sit better in the mix. The other snare has similar plug-ins, as well as the Waves L1 [limiter] and the Lexicon Room [reverb], set to a Tiled Room, with a splashy sound, to give a bit more power at the end of the song.”
D’Armancourt: “We had the SoundToys Radiator [saturation plug-in] on the bass DI and Justin and Chris’s acoustic guitar tracks to add some dirt and harmonics. Basically our problem was that we were mixing material coming from high-quality studio microphones, boom microphones, lavalier microphones and DI, and if we did not process things in certain ways, we’d have ended up with several different layers that don’t really fit together. So throughout the whole mix process we were very focused on making sure everything blended together and that it sounded coherent. We used harmonics generators like the Radiator and Devil-Loc to this end.”
Talking about the bass, one area in which d’Armancourt and de la Villéon’s mix of ‘Say Something’ is notably different from the urban and pop music that’s out there is that it’s relatively bass-light. Apparently there was a good reason for this. D’Armancourt: “If the first minute of the video, when the studio samples and pre-recorded tracks can be heard, is amazing sound-wise, then the rest of the video could become disappointing, as we could never match the impact of this pre-recorded material. So instead we opted to build throughout the video, so that the viewer is engaged throughout and can still be amazed until the end.”
De la Villéon: “We also did not have a lot of bass to work with. There’s only the bass guitar, there are no synths with low end, and the kick sound actually was not that low. When the Maschine plays in the beginning, we also wanted to keep it small because it’s in a room, and not yet in the big hall.”
Justin Timberlake’s lead vocals consist of three tracks, which, says d’Armancourt, “all come from one track of lavalier recording. The two boom mics that also capture some of his vocals are further down in the session. But many other things were picked up by these boom mics as well, which is why we did not put them with the vocal tracks. We split Justin’s vocal recording over three tracks to make it easier to treat each track differently. The ‘JT Intro’ track is where you see him with the NI Maschine, and our idea there was that we wanted to bring out the physical sound of him actually hitting the pads, because we really liked that. It brings that controller into the analogue world. The SPL Transient Designer and the FabFilter Pro-G gate help to bring these pad hits out.”
De la Villéon: “The main ‘JT’ and the ‘JT Verse’ tracks are pretty much the same, only the ‘Verse’ track has Justin’s vocals without the guitar, which led to different EQ settings on the Pro-Q2. The Pro-Q2 is most of all a corrective EQ, because the sound of lavalier mics has a lot of 5-600 Hz and also a lot of what you don’t want to hear, depending on where you place it on someone. We use the [UAD] Maag EQ on all vocal tracks to alter the sound more to our taste, particularly applying the Air band function of that EQ, which sounds really great. We also removed some 2.5kHz, with a large Q, to get the vocal to sound smoother. Then there’s the Softube Tube-Tech CL1B compressor, and then the Melda Dynamic EQ, which is pretty cool if you want to take off some harshness, and lavalier mics can sound a bit harsh. In this case we use it to add some high end during quiet parts and it cuts high end during louder sections to avoid harshness.
“We also really wanted to have a small room sound during the intro, so the verse track has a send to the ‘Short Room’ aux track with the Altiverb small space reverb. When Justin enters the elevator, we take the reverbs down to zero, and you get the sense of the sound becoming small and boxy, and then we push up big reverbs as he enters the main building, using both the Altiverb and HOFA IQ reverbs. There’s also panning automation on the latter, which is on the moment Justin appears at the left side of the screen, when they sing harmonies a cappella. We really wanted to have Justin and Chris very separate at that point, because the camera is far away from Chris, and yet the focus is on him, with Justin blurred on the left.”
D’Armancourt: “The choir aux track has the SPL Transient Designer, to shorten the natural reverb, the SSL compressor, and the Flux Stereo Tool. It’s one of Flux’s rare pieces of freeware, and it’s really useful to narrow or widen stuff. It doesn’t just work with M-S processing, but also does weird stuff with phasing. We also have the FabFilter Pro-DS [de-esser] on some of the ambient mics, because at the end the drummer hits his cymbals pretty hard, and that became overwhelming. In a case like this we find the Pro-DS handy to find the frequency and reduce it.”
The master bus in this session is very sparsely populated by modern standards, with just the SoundToys Sie-Q EQ and the Waves WLM Loudness meter. De la Villéon: “We used the EQ on the master bus only very lightly, bringing a tiny bit more life to the mix by adding just 1dB of 3.5kHz. The loudness meter is there because we knew the video would end up on Justin’s YouTube channel, so we had to care about this, and make sure it was loud enough. At the same time we didn’t want it to be too hot and reduce the dynamics. When we were in Jungle City Studios we routed the session through the SSL Duality, which added a bit of an analogue quality to the mix.
“Afterwards the track was mastered. We had some discussions with Justin about this, and decided to work with Chab in Paris, who won a Grammy for his work on mastering Daft Punk’s Random Access Memories. Mastering was important, because the sound of this video is quite different to what you usually hear on YouTube music videos. There’s lots of ambience without a strong centre, and we wanted to make sure the ambience wasn’t reduced. We’re super happy with what he did.”
The French duo were, naturally, equally happy with the enthusiastic response their efforts received from Timberlake and his team, while stressing again that they only could achieve this because of the “invisible team” behind them at La Blogothèque — where, says d’Armancourt, “a whole team of engineers have worked for 10 years to improve all these techniques”. The video itself has enjoyed 47 million YouTube views at the time of writing, and played its part in pushing the album version of ‘Say Something’ to the top 10 in the US, and Timberlake’s Man Of The Woods album to number two in the UK and the US top spot.
The decision that the sound mix should be more polished and studio-like than that of typical Blogothèque mixes meant that Henri d’Armancourt and Guillaume de la Villéon used some techniques not normally employed in their other videos. De la Villéon: “Despite the rule of the Takeaway Shows that everything you see is in the soundtrack, and vice versa, because it had to go on Justin ’s channel and appeal to his fans, we felt quite free to use different post-production tricks to enhance what we had recorded with the microphones. And the vocals are always the weakest signal to deal with, yet they’re also the most important. So post-production treatments are almost inevitable.”
D’Armancourt: “We try not to use artificial reverb in the Takeaway Shows. But we had so much spill on the ambient mics that we could not extract a useful vocal reverb from them. The ambient mics were the most useful for reverb on loud things, like the electric guitar or the drums. Sometimes also we used artificial reverb because the sound of the lavalier microphones is not particularly natural. It does give you something really present and up close, like you have your ear against someone’s chest. But at the same time you need to add some reverb so it does not sound claustrophobic.
“The question we always asked ourselves was: ‘Does it sound good?’ And then: ‘Does it sound good with the image?’ Those were the two golden rules. In film production, when you are sound editing or are doing re-recording mixing, you always try to put yourself in the mind-set of the viewer, and often this involves doing things that are not ‘real’. For example, they often add footsteps in the movies, but in real life, unless someone is walking in high heels on a concrete floor, you don’t really hear footsteps that much. So Justin and Chris walking around was not an issue when we mixed. On the other hand, the metallic elevator sounds are pretty present, and we could not take them out. But we were quite happy with these sounds, because they added to the sense of being in the Bradbury Building.”
De la Villéon: “If you have footage of someone in the Grand Canyon, you imagine what it might sound like, but the reality may be very different. Similarly, we knew what the Bradbury Building sounded like, but when we added reverb during mixing we did not necessarily try to copy that exactly, but instead we went more for what fitted the images and what people would imagine the building would sound like. So we did not try to measure the reverb in the building. We knew that the decay was around 3.5 seconds and we were aware of the frequencies in the reverb. But we did not feel that we had to match that exactly.”
Guillaume de la Villéon began playing violin aged four, and also played bass and drums before going more deeply into studio work. He cut his engineering teeth as an assistant at Studio Pigalle in Paris, and then switched to live sound. In addition to having freelanced for La Blogothèque for eight years, he regularly tours with well-known French bands as their front-of-house engineer. To this end he has his own mobile studio. He prides himself in still recording and mixing on average three studio albums a year, and thereby being able to keep his feet in both live sound and studio camps.
Henri d’Armancourt plays bass, guitar and sings, and is the singer and guitarist in his band Shoefiti. While still at school he was an assistant at Studio Sequenza in Montreuil, near Paris, which records mostly world, jazz and classical music. After leaving school, d’Armancourt began working in film and on TV shows as the production sound mixer, and sometimes a boom operator. He began doing freelance work for La Blogothèque six years ago, and complements this with work as a sound engineer, mixer and producer. Like de la Villéon he prides himself in being able to work in both studio, live band and AV situations, calling himself “kind of like a Swiss Army knife!”
Henri d’Armancourt: “Justin’s team really liked the piezo [under-saddle pickup] sound of the acoustic guitars, so we used them for his and Chris’s guitar, but over the last 10 years we have at the Blogothèque developed a technique of placing a DPA 4060 inside acoustic guitars. We place the mic on the strut inside the soundhole, just in front of the neck, and we then run the wire outside the guitar alongside the fretboard, just below it, so it’s in the shadow of the fretboard, and then to the back of the guitar. For us, using a mic results in a warmer and more natural guitar sound. Elliot Ives [the other guitarist in the video] allowed us to install the DPA 4060 inside of his acoustic guitar.”