Baby Driver

You don’t have to be a rabid fan of Edgar Wright’s work to know of his films. His comedy trilogy (Shaun of the Dead, Hot Fuzz, The World’s End) and cult classics like Scott Pilgrim vs. the World loom large in pop culture. His films have earned a life beyond most films’ brief release period and earned Wright a loyal following. The latest film from Wright is Baby Driver, a musically-fueled action film written and directed by Wright, which just made a big splash at SXSW. It stars Ansel Elgort, Kevin Spacey, Jon Hamm, Jamie Foxx, and Eiza Gonzalez.

At NAB, Avid brought in a number of featured speakers for its main stage presentations, as well as its Avid Connect event. One of these speakers was Paul Machliss (Scott Pilgrim vs. the World, The World’s End, Baby Driver), who spoke to packed audiences about the art of editing these films. I had a chance to go in-depth with Machliss about the complex process of working on Baby Driver.

From Smoke to baptism by fire

We started our conversation with a bit of the backstory of the connection between Wright and Machliss. He says, “I started editing as an online editor and progressed from tape-based systems to being one of the early London-based Smoke editors. My boss at the time passed along a project that he thought would be perfect for Smoke. That was onlining the sitcom Spaced, directed by Edgar Wright. Edgar and I got on well. Concurrent to that, I had started learning Avid. I started doing offline editing jobs for other directors and had a ball. A chance came along to do a David Beckham documentary, so I took the plunge from being a full-time online editor to taking my chances in the freelance world. On the tail end of the documentary, I got a call from Edgar, offering me the gig to be the offline editor for the second season of Spaced, because Chris Dickens (Hot Fuzz, Berberian Sound Studio, Slumdog Millionaire) wasn’t available to complete the edit. And that was really jumping into the deep end. It was fantastic to be able to work with Edgar at that level.”

Machliss continues, “Chris came back to work with Edgar on Shaun of the Dead and Hot Fuzz, so over the following years I honed my skills working on a number of British comedies and dramas. After Slumdog Millionaire came out, which Chris cut and for which he won a number of awards, including an Oscar, Chris suddenly found himself very busy, so the rest of us working with Edgar all moved up one in the queue, so to speak. The opportunity to edit Scott Pilgrim came up, so we all threw ourselves into the world of feature films, which was definitely a baptism by fire. We were very lucky to be able to work on a project of that nature during a time where the industry was in a bit of a slump due to the recession. And it’s fantastic that people still remember it and talk about it seven years on. Which brings us to Baby Driver. It’s great when a studio is willing to invest in a film that isn’t a franchise, a sequel, or a reboot.”

Music drives the film

In Baby Driver, Ansel Elgort plays “Baby”, a young kid who is the getaway driver for a gang. At a young age, he was in a car accident which leaves him with tinnitus, so it takes listening to music 24/7 to drown out the tinnitus. Machliss explains, “His whole life becomes regimented to whatever music he is listening to – different music for different moods or occasions. Somehow everything falls magically into sync with whatever he is listening to – when he’s driving, swerving to avoid a car, making a turn – it all seems to happen on the beat. Music drives every single scene. Edgar deliberately chose commercial top-20 tracks from the 1960s up to today. Each song Baby listens to also slyly comments on whatever is happening at the time in the story. Everything is seemingly choreographed to musical rhythms. You’re not looking at a musical, but everything is musically driven.”

Naturally, building a film to popular music brings up a whole host of production issues. Machliss tells how this film had been in the planning for years, “Edgar had chosen these tracks years ago. I believe it was in 2011 that Edgar and I tried to sequence the tracks and intersperse them with sound effects. A couple of months later, he did a table read in LA and sent me the sound files. In the Avid, I combined the sound files, songs, and some sound effects to create effectively a 100-minute radio play, which was, in fact, the film in audio form. The big thing is that we had to clear every song before we could start filming. Eventually we cleared 30-odd songs for the film. In addition, Edgar worked with his stunt team and editor Evan Schiff in LA to create storyboards and animatics for all of the action scenes.”

Editor on the front lines

Unlike most films, a significant amount of the editing took place on-set with Machliss working from a portable set-up. He says, “Based on our experiences with Scott Pilgrim and World’s End, Edgar decided it would be best to have me on-set during most of the Atlanta shoot for Baby Driver. Even though a cutting room was available, I was in there maybe ten percent of the time. The rest of the time I was on set. I had a trolley with a laptop, monitor, an Avid Mojo, and some hard drives and I would connect myself via ethernet to the video assist’s hard drive. Effectively I was crew in the front lines with everyone else. Making sure the edit worked was as important as getting a good take in the can. If I assured Edgar that a take would work, then he knew it wasn’t going to come back and cause problems for us six months later. We wanted things to work naturally in camera without a lot of fiddling in post. We didn’t want to have to fall back on frame-cutting and vari-speeding if we didn’t have to. There was a lot of prep work in making sure actions correctly coincided with certain lyrics without the action seeming mechanical.”

The nature of the production added to the complexity of the production audio configuration, too. Machliss explains, “Sound-wise, it was very complicated. We had playback going to earwigs in the actors’ ears, Edgar wanted to hear music plus the dialogue in his cans, and then I needed to get a split feed of the audio, since I already had the clean music on my timeline. We shot this mostly on 35mm film. Some days were A-camera only, but usually two cameras running. It was a combination of Panavision, Arricams, and occasionally Arri Alexas. Sometimes there were some stunt shots, which required nine or ten cameras running. Since the action all happened against playback of a track, this allowed me to use Avid’s multicam tools to quickly group shots together. Avid’s AMA tools have really come of age, so I was able to work without needing to ingest anything. I could treat the video assist’s hard drive as my source media, as long as I had the ethernet connection to it. If we were between set-ups, I could get Avid to background-transcode the media, so I’d have my own copy.”

Did all of this on-set editing speed up the rest of the post process? He continues, “All of the on-set editing helped a great deal, because we went into the real post-production phase knowing that all the sequences basically worked. During that time, as I’d fill up a LaCie Rugged drive, I would send that back to the suites. My assistant, Jerry Ramsbottom, would then patiently overcut my edits from the video assist with the actual scanned telecine footage as it came in. We shot from mid-February until mid-May and then returned to England. Jonathan Amos came on board a few weeks into the director’s cut edit and worked on the film with Edgar and myself up until the director’s cut picture lock. He did a pass on some of the action scenes while Edgar and myself concentrated on dialogue and the overall shape of the film. He stayed on board up until the final picture lock and made an incredible contribution to the action and the tension of the film. By the end of the year we’d locked and then we finished the final mix mid-February of this year. But the great thing was to be able to come into the edit and have those sequences ready to go.”

Editing from set is something many editors try to avoid. They feel they can be more objective that way. Machliss sees it a bit differently, “Some editors don’t like being on set, but I like the openness of it – taking it all in. Because when you are in the edit, you can recall the events of the day a particular scene was shot – ‘I can remember when Kevin Spacey did this thing on the third take, which could be useful’. It’s not vital to work like this, but it does preclude to a kind of short-hand, which is something Edgar and I have developed over these years anyway. The beauty of it is that Edgar and I will take the time to try every option. You can never hit on the perfect cut the first time. Often you’ll get feedback from screenings, such as ‘we’d like to see more emotion between these characters’. You know what’s available and sometimes four extra shots can make all the difference in how a scene reads without having to re-imagine anything. We did drop some scenes from the final version of the film. Of course, you go ‘that’s a shame’, but at least these scenes were given a chance. However, there are always bits where upon the 200th viewing you can decide, ‘well, that’s completely redundant’ – and it’s easy to drop. You always skate as close to the edge of making a film shorter without doing any damage to it.”

The challenge of sound

During sound post, Baby Driver also presented some unique challenges. Machliss says, “For the sound mix – and even for the shoot – we had to make sure we were working with the final masters of the song recordings to make sure the pitch and duration remained constant throughout. Typically these came in as mono or stereo WAVs. Because music is such an important element to the film, the concept of perceived direction becomes important. Is the music emanating from Baby’s earbuds? What happens to it when the camera moves or he turns his head? We had to work out a language for the perception of sound. This was Edgar’s first film mixed in Dolby ATMOS and we were the second film in Goldcrest London’s new Atmos-certified dubbing theater. Then we did a reduction to 7.1 and 5.1. Initially we were thinking this film would have no score other than the songs. Invariably you need something to get from A to B. We called on the services of Steven Price (Gravity, Fury, Suicide Squad), who provided us with some original cues and some musical textures. He did a very clever thing where he would match the end pitch or notes of a commercial song and then by the time he came to the end of his cue, it would match to the incoming note or key of the next song. And you never notice the change.”

Working with Avid in a new way

To wrap up the conversation, we talked a bit about using Avid Media Composer on his work. Machliss has used numerous other systems, but Media Composer still fits the bill for his work today. He says, “For me, the speed of working with AMA in Avid in the latest software was a real benefit. I could actually keep up with the speed of the shoot. You don’t want to be the one holding up a crew of 70. I also made good use of background transcoding. On a different project (Fleabag), I was able to work with native 2K Alexa ProRes camera files at full resolution. It was fantastic to be able to use Frameflex and apply LUTs – doing the cutting, but then bringing back my old skills as an online editor to paint out booms and fix things up. Once we locked, I could remove the LUTs and export DPX files, which went straight to the grading facility. That was exciting to work in a new way.”

Baby Driver opened at the start of July in the US and is a fun ride. You can certainly enjoy a film like this without knowing the nitty gritty of the production that goes into it. However, after you’ve read this article, you just might need to see it at least twice – once to just enjoy and once again to study the “invisible art” that’s gone into bringing it to screen.

(For more with Paul Machliss, check out these interviews at Studio Daily, ProVideoCoalition, and FrameIO.)

Originally written for Digital Video magazine / Creative Planet Network

©2017 Oliver Peters

Bricklayers and Sculptors

One of the livelier hangouts on the internet for editors to kick around their thoughts is the Creative COW’s Apple Final Cut Pro X Debates forum. Part forum, part bar room brawl, it started as a place to discuss the relative merits (or not) of Apple’s FCP X. As such, the COW’s bosses allow a bit more latitude than in other forums. However, often threads derail into really thoughtful discussions about editing concepts.

Recently one of its frequent contributors, Simon Ubsdell, posted a thread called Bricklayers and Sculptors. In his words, “There are two different types of editors: Those who lay one shot after another like a bricklayer builds a wall. And those who discover the shape of their film by sculpting the raw material like a sculptor works with clay. These processes are not the same. There is no continuum that links these two approaches. They are diametrically opposed.”

Simon Ubsdell is the creative director, partner, and editor/mixer for London-based trailer shop Tokyo Productions. Ubsdell is also an experienced plug-in developer, having developed and/or co-developed the TKY, Tokyo, and Hawaiki effects plug-ins. But beyond that, Simon is one of the folks with whom I often have e-mail discussions regarding the state of editing today. We were both early adopters of FCP X who have since shifted almost completely to Adobe Premiere Pro. In keeping with the theme of his forum post, I asked him to share his ideas about how to organize an edit.

With Simon’s permission, the following are his thoughts on how best to organize editing projects in a way that keeps you immersed in the material and results in editing with greater assurance that you’ve make the best possible edit decisions.

________________________________________________

Simon Ubsdell – Bricklayers and Sculptors in practical terms

To avoid getting too general about this, let me describe a job I did this week. The producer came to us with a documentary that’s still shooting and only roughly “edited” into a very loose assembly – it’s the stories of five different women that will eventually be interweaved, but that hasn’t happened yet. As I say, extremely rough and unformed.

I grabbed all the source material and put it on a timeline. That showed me at a glance that there was about four hours of it in total. I put in markers to show where each woman’s material started and ended, which allowed me to see how much material I had for each of them. If I ever needed to go back to “everything”, it would make searching easier. (Not an essential step by any means.)

I duplicated that sequence five times to make sequences of all the material for each woman. Then I made duplicates of those duplicates and began removing everything I didn’t want. (At this point I am only looking for dialogue and “key sound”, not pictures which I will pick up in a separate set of passes.)

Working subtractively

From this point on I am working almost exclusively subtractively. A lot of people approach string-outs by adding clips from the browser – but here all my clips are already on the timeline and I am taking away anything I don’t want. This is for me the key part of the process because each edit is not a rough approximation, but a very precise “topping and tailing” of what I want to use. If you’re “editing in the Browser” (or in Bins), you’re simply not going to be making the kind of frame accurate edits that I am making every single time with this method.

The point to grasp here is that instead of “making bricks” for use later on, I am already editing in the strictest sense – making cuts that will stand up later on. I don’t have to select and then trim – I am doing both operations at the same time. I have my editing hat on, not an organizing hat. I am focused on a timeline that is going to form the basis of the final edit. I am already thinking editorially (in the sense of creative timeline-based editing) and not wasting any time merely thinking organizationally.

I should mention here that this is an iterative process – not just one pass through the material, but several. At certain points I will keep duplicates as I start to work on shorter versions. I won’t generally keep that many duplicates – usually just an intermediate “long version”, which has lost all the material I definitely don’t want. And by “definitely don’t want” I’m not talking about heads and tails that everybody throws away where the camera is being turned on or off or the crew are in shot – I am already making deep, fine-grained editorial and editing decisions that will be of immense value later on. I’m going straight to the edit point that I know I’ll want for my finished show. It’s not a provisional edit point – it’s a genuine editorial choice. From this point of view, the process of rejecting slates and tails is entirely irrelevant and pointless – a whole process that I sidestep entirely. I am cutting from one bit that I want to keep directly to the next bit I want to keep and I am doing so with fine-tuned precision. And because I am working subtractively I am actually incorporating several edit decisions in one – in other words, with one delete step I am both removing the tail from the outgoing clip and setting the start of the next clip.

Feeling the pacing and flow

Another key element here is that I can see how one clip flows into another – even if I am not going to be using those two clips side-by-side. I can already get a feel for the pacing. I can also start to see what might go where, so as part of this phase, I am moving things around as options start suggesting themselves. Because I am working in the timeline with actual edited material, those options present themselves very naturally – I’m getting offered creative choices for free. I can’t stress too strongly how relevant this part is. If I were simply sorting through material in a Browser/Bin, this process would not be happening or at least not happening in anything like the same way. The ability to reorder clips as the thought occurs to me and for this to be an actual editorial decision on a timeline is an incredibly useful thing and again a great timesaver. I don’t have to think about editorial decisions twice.

And another major benefit that is simply not available to Browser/Bin-based methods, is that I am constructing editorial chunks as I go. I’m taking this section from Clip A and putting it side-by-side with this other section from Clip A, which may come from earlier in the actual source, and perhaps adding a section from Clip B to the end and something from Clip C to the front. I am forming editorial units as I work through the material. And these are units that I can later use wholesale.

Another interesting spin-off is that I can very quickly spot “duplicate material”, by which I mean instances where the same information or sentiment is conveyed in more or less the same terms at different places in the source material. Because I am reviewing all of this on the timeline and because I am doing so iteratively, I can very quickly form an opinion as to which of the “duplicates” I want to use in my final edit.

Working towards the delivery target

Let’s step back and look at a further benefit of this method. Whatever your final film is, it will have the length that it needs to be – unless you’re Andy Warhol. You’re delivering a documentary for broadcast or theatrical distribution, or a short form promo or a trailer or TV spot. In each case you have a rough idea of what final length you need to arrive at. In my case, I knew that the piece needed to be around three minutes long. And that, of course, throws up a very obvious piece of arithmetic that it helps me to know. I had five stories to fit into those three minutes, which meant that the absolute maximum of dialogue that I would need would be just over 30 seconds from each story!  The best way of getting to those 30 seconds is obviously subtractively.

I know I need to get my timeline of each story down to something approaching this length. Because I’m not simply topping and tailing clips in the Browser, but actually sculpting them on the timeline (and forming them into editorial units, as described above), I can keep a very close eye on how this is coming along for each story strand. I have a continuous read-out of how well I am getting on with reducing the material down to the target length. By contrast, if I approach my final edit with 30 minutes of loosely selected source material to juggle, I’m going to spend a lot more time on editorial decisions that I could have successfully made earlier.

So the final stage of the process in this case was simply to combine and rearrange the pre-edited timelines into a final timeline – a process that is now incredibly fast and a lot of fun. I’ve narrowed the range of choices right down to the necessary minimum. A great deal of the editing has literally already been done, because I’ve been editing from the very first moment that I laid all the material on the original timeline containing all the source material for the project.

As you can see, the process has been essentially entirely subtractive throughout – a gradual whittling down of the four hours to something closer to three minutes. This is not to say there won’t be additive parts to the overall edit. Of course, I added music, SFX, and graphics, but from the perspective of the process as a whole, this is addition at the most trivial level.

Learning to tell the story in pictures

There is another layer of addition that I have left out and that’s what happens with the pictures. So far I’ve only mentioned what is happening with what is sometimes called the “radio edit”. In my case, I will perform the exact same (sometimes iterative) process of subtracting the shots I want to keep from the entirety of the source material – again, this is obviously happening on a timeline or timelines. The real delight of this method is to review all the “pictures” without reference to the sound, because in doing so you can get a real insight into how the story can be told pictorially. I will often review the pictures having very, very roughly laid up some of the music tracks that I have planned on using. It’s amazing how this lets you gauge both whether your music suits the material and conversely whether the pictures are the right ones for the way you are planning to tell the story.

This brings to me a key point I would make about how I personally work with this method and that’s that I plunge in and experiment even at the early stages of the project. For me, the key thing is to start to get a feel for how it’s all going to come together. This loose experimentation is a great way of approaching that. At some point in the experimentation something clicks and you can see the whole shape or at the very least get a feeling for what it’s all going to look like. The sooner that click happens, the better you can work, because now you are not simply randomly sorting material, you are working towards a picture you have in your head. For me, that’s the biggest benefit of working in the timeline from the very beginning. You’re getting immersed in the shape of the material rather than just its content and the immersion is what sparks the ideas. I’m not invoking some magical thinking here – I’m just talking about a method that’s proven itself time and time again to be the best and fastest way to unlock the doors of the edit.

Another benefit is that although one would expect this method to make it harder to collaborate, in fact the reverse is the case if each editor is conversant with the technique. You’re handing over vastly more useful creative edit information with this process than you could by any other means. What you’re effectively doing is “showing your workings” and not just handing over some versions. It means that the editor taking over from you can easily backtrack through your work and find new stuff and see the ideas that you didn’t end up including in the version(s) that you handed over. It’s an incredibly fast way for the new editor to get up to speed with the project without having to start from scratch by acquainting him or herself with where the useful material can be found.

Even on a more conventional level, I personally would far rather receive string-outs of selects than all the most carefully organized Browser/Bin info you care to throw at me. Obviously if I’m cutting a feature, I want to be able to find 323T14 instantly, but beyond that most basic level, I have no interest in digging through bins or keyword collections or whatever else you might be using, as that’s just going to slow me down.

Freeing yourself of the Browser/Bins

Another observation about this method is how it relates to the NLE interface. When I’m working with my string-outs, which is essentially 90% of the time, I am not ever looking at the Browser/Bins. Accordingly, in Premiere Pro or Final Cut Pro X, I can fully close down the Project/Browser windows/panes and avail myself of the extra screen real estate that gives me, which is not inconsiderable. The consequence of that is to make the timeline experience even more immersive and that’s exactly what I want. I want to be immersed in the details of what I’m doing in the timeline and I have no interest in any other distractions. Conversely, having to keep going back to Bins/Browser means shifting the focus of attention away from my work and breaking the all-important “flow” factor. I just don’t want any distractions from the fundamentally crucial process of moving from one clip to another in a timeline context. As soon as I am dragged away from that, there’s is a discontinuity in what I am doing.

The edit comes to shape organically

I find that there comes a point, if you work this way, when the subsequence you are working on organically starts to take on the shape of the finished edit and it’s something that happens without you having to consciously make it happen. It’s the method doing the work for you. This means that I never find myself starting a fresh sequence and adding to it from the subsequences and I think that has huge advantages. It reinforces my point that you are editing from the very first moment when you lay all your source material onto one timeline. That process leads without pause or interruption to the final edit through the gradual iterative subtraction.

I talked about how the iterative sifting process lets you see “duplicates”, that’s to say instances where the same idea is repeated in an alternative form – and that it helps you make the choice between the different options. Another aspect of this is that it helps you to identify what is strong and what is not so strong. If I were cutting corporates or skate videos this might be different, but for what I do, I need to be able to isolate the key “moments” in my material and find ways to promote those and make them work as powerfully as possible.

In a completely literal sense, when you’re cutting promos and trailers, you want to create an emotional, visceral connection to the material in the audience. You want to make them laugh or cry, you want to make them hold their breath in anticipation, or gasp in astonishment. You need to know how to craft the moments that will elicit the response you are looking for. I find that this method really helps me identify where those moments are going to come from and how to structure everything around them so as to build them as strongly as possible. The iterative sifting method means you can be very sure of what to go for and in what context it’s going to work the best. In other words, I keep coming back to the realization that this method is doing a lot of the creative work for you in a way that simply won’t happen with the alternatives. Even setting aside the manifest efficiency, it would be worth it for this alone.

There’s a huge amount more that I could say about this process, but I’ll leave it there for now. I’m not saying this method works equally well for all types of projects. It’s perhaps less suited to scripted drama, for instance, but even there it can work effectively with certain modifications. Like every method, every editor wants to tweak it to their own taste and inclinations. The one thing I have found to its advantage above all others is that it almost entirely circumvents the problem of “what shot do I lay down next?” Time and again I’ve seen Browser/Bin-focused editors get stuck in exactly this way and it can be a very real block.

– Simon Ubsdell

For an expanded version of this concept, check out Simon’s in-depth article at Creative COW. Click here to link.

For more creative editing tips, click on this link for Film Editor Techniques.

©2017 Simon Ubsdell, Oliver Peters