The Missing Details of Baby Driver’s Crazy Workflow
Edgar Wright is a filmmaker’s filmmaker with a quirky style that is instantly recognizable. But his latest and most commercially successful film to date, Baby Driver, takes things up a notch.
Wright’s highly-unusual shooting process created some unique workflow challenges for Paul Machliss, ACE, its lead editor, who described it as “the most challenging film he’s ever worked on.”
Much has already been written about Baby Driver, but I was lucky enough to have a 1-on-1 conversation with Paul and get insights I haven’t read elsewhere.
While most 35mm film shoot workflows are
Shoot > Scan > Edit
Baby Driver’s workflow was more like
Edit > Shoot > Edit > Scan > Edit.
Herein lie the details of how the images went from animatics, to 35mm film with real-time on-set editing, to a digital intermediate scan, to NLE, to color grade, to projection at your local cineplex.
What a ride. So strap in, and hang on.
To get started, take a look the opening car chase from the movie.
The Editing Starts in Pre-production
Most directors do a storyboard, but Edgar, Paul and L.A.-based editor Evan Schiff did a “pre-edit” of the film ahead of the shoot using Animatics (animated storyboards). Paul brought the animatics into Avid and painstakingly cut them beat-by-beat to line up perfectly with the soundtrack which Edgar literally spent years ahead of time selecting.
A lot of what makes Baby Driver so special is how the editing, action, and cinematography seem to be meticulously connected and synched. The timing of each scene was pre-planned to a level that’s unheard-of for an action film, and this pre-edit was vital as a guide to the timing on set.
Paul was on set for most of the shoot and I wanted to know why he specifically needed to be there as the editor. Isn’t that the assistant camera or continuity person’s job, to make sure the shot looks good after the director calls “cut”?
“It grew out of the way Edgar and I were working on set from Scott Pilgrim Vs the World (2010) to The World’s End (2013) where we were working on a lot of action scenes.” Explains Paul. “The idea isn’t for me just to be on set and say, ‘Yeah, that looks like a good shot.’ Because you’re right. There’s plenty of films, let’s say 99.9% of them, that get made without the editor on set; and there are plenty of people to say the shot looks good.
“But, there’s a difference between saying ‘The shot looks good’ and ‘The shot works’.
“The shot has to work within the boundaries of the storytelling. And we set ourselves a particular criteria here.”
Choreographing Dialogue Scenes and Car Chases
This was the crux of Paul’s purpose on set. Not just to make sure they got the shot, but to ensure that the action in the scene, whether it was a car chase or a conversation, fit within the context of the music and choreography.
That’s right, choreography. You might think of a traditional musical as having “choreography.” But an action flick? Yep.
Many of the scenes were choreographed to fit to the music. (Ryan Heffington is credited as the choreographer on the film.) In addition to the car chases, when, where, and how the actors proceeded through any scene with music was also choreographed. The actors even had the music piped to them via ear pieces.
“People might think, ‘Why do you need a choreographer?’ All Ansel is doing is walking across to that part of the kitchen, picking up a carton of milk and putting it down on the bench. [Ansel Elgort plays the main character, Baby] Why do you need to have someone count that in?’ Well, you do because it all ties in with the music. So there was a lot of 5-6-7-8’s going on on set to make sure everyone was getting it on the beat.
“If you’re filming a shot, and the time it takes for a certain bit of action to happen on that shot is actually a second and a half longer than the song or the music will allow, what you can’t really do is go, ‘Oh, I’ll just make the music a little bit longer in the end.’ If someone is singing a verse, you can’t just magically invent a second and a half to solve your filming problem. You have committed to this particular song and by the end of this line in the second verse, the actor needs to have done, this, this, this, this and this. What we really didn’t want to do was go back to the edit and start pulling out frames, or do vari-speeding, or a lot of particular tricks.”
When you’re working to have your actors and actions so meticulously choreographed, you can’t leave anything to chance. So, at the end of each take, Paul would drop the shot into the Avid timeline, on a track above the pre-edited animatics, to make sure the continuity and duration worked.
Paul was editing the film quite literally in real time as the shots came in, replacing them one at a time.
The Film to Digital Workflow
What makes all of this even more fascinating is that 90 to 95% of the film was shot on 35mm film. Edgar Wright is one of the lasting holdouts on using film vs. digital. (They did use the Arri Alexa for some of the night shots, where the Alexa had a better latitude than the 35mm; and they used the Alexa Mini for some of the shots where there were tricky corners.)
I took this opportunity to ask something I’ve wondered for a long time.
“So if you’re shooting in 35mm, then is the video you’re dropping into Media Composer from the video assist unit?”
“Yes. We’re making a 21st-century film with very 20th-century methods. A good old fashioned Panavision 35mm camera, or the Arricam ST. The video tap would get recorded by the video assist, using the QTake software. And he would record them as ProRes files. I would be tethered to him via an ethernet cable. I wouldn’t be taking in or recording video files. We’d be creating a network basically. The hard drive connected to his Mac Pro I could effectively use as my source drive.
“As soon as Edgar called ‘Cut!’ and the video assist stopped recording, I would immediately see a QuickTime file in the finder. Then using the Avid AMA function [Avid Media Access] which has the ProRes codec built in, I could immediately drag the clip from the Finder into the timeline. In that method I generated a lot of media because then I would use the background transcode to then make my own copy of it and relink to that; because, of course, as soon as the video assist powered down, I would lose all the media via the AMA link.”
Sounds like a piece of cake. 🙂
This process would generate a lot of media that Paul saved on an 8TB Thunderbolt drive which he had with him the entire time. But that media was comprised of files from the video tap of the 35mm film camera, captured by QTake on the video assist, which was then transcoded by Paul’s Avid into a DNx36 HD Avid media.
Paul is now building his timeline (or, shall we say, replacing his timeline) with these DNx36 proxy files. But no matter the quality of the codec he’s using, this is only video assist footage, which looks nothing like the images that are simultaneously being captured on the physical film.
So the next question is, how do you connect that timeline with the actual film? Daily footage would be flown from Atlanta (where they were shooting) to Fotokem in Los Angeles, where it was processed and developed. It was then sent to eFilm to be scanned to digital intermediate files.
The Weekly Grading Process
Another unusual aspect of this film was how closely Paul worked with not only Edgar, but also Bill Pope, Edgar’s go-to DP (or DoP as they say in the UK).
“There’s something very nice actually about the DoP and the editor being able to have a constant dialog, as well. We’re always talking on set, and he’s always coming over to review stuff.”
Once a week, he, Edgar and Bill would go to the Atlanta offices of Company 3 to watch digital selects streamed from Los Angeles, and do a preliminary grade. Those grading notes would be sent back to eFilm, where they could begin a more detailed grading process on the rushes. So by the time the digitized rushes were sent back to Paul for editing (usually via Aspera), the film was already pretty close to the look that Edgar and Bill desired.
It could be 2-3 days before they’d get the digitized footage back (In the meantime, Paul is continuing his on-set cutting adventure on the next day’s shoot). Once the film scans were received, his first assistant editor, Jerry Ramsbottom, would ingest the digitized footage, then relink to the Avid Media footage (the DNx36 footage from video assist).
To facilitate this process, Paul would add as much metadata as possible to the transcoded footage. But the only real link was the audio timecode (specifically Time of Day timecode from the video assist which was being generated by the sound recordist). The audio playback guy (who handled the music playback on set) would also send Paul the timecode from the BWAV files from the music tracks. Jerry would use all this timecode to relink the lower quality proxy media files to the pristine, high-resolution telecine scans, and it would work perfectly about 95% of the time using this method.
Working with Co-Editor Jon Amos
Paul would assemble as much as the film as possible in Atlanta since he was on set 90% of the time. During the last week of shooting, most of the media would be shipped to London, where an editorial team would start the work of fine editing, lead by the co-editor on the project Jon Amos. It was Jon’s job to go through all the big action scenes (which Paul says he’s particularly brilliant at editing). Jon would also work on fixing those parts of the edit that Paul didn’t have time to correct on set, or ran into creative obstacles. For instance, in spite of all of their efforts, there were some times when the amount of footage for a scene shot was longer than the music allowed, so Jon would fix those as well.
The Archival, Retrieval and Distribution Process
I just had to chime in and ask, “So since you’re editing digitally, when it’s all done, do they take the final digital master you’ve edited and cut the original film based on the digital?”
“No. The procedure nowadays is that film is effectively just the acquisition medium. The negative gets scanned, usually at 2K, but increasingly also at 4K resolution. The film then actually gets put in a vault. And you almost never touch it again. The film basically gets converted to digital files and converted LTO media, and that becomes your main source for the final conform.
“Sometimes you have to go back and extract some extra footage. There were a couple of instances in Baby Driver where it was scanned at 2K, but I needed to go in about 50 or 60% to reframe; and at that point, 2K wouldn’t have coped with it. So you basically specify the frame range, then the rolls get pulled out and they rescan just the frames you need at 4K.”
Paul went on to explain that all future footage requests from other departments (e.g. VFX) will get the footage they need from those master LTOs. Unless there is some need to rescan, the original film footage will not be touched again. The entire post production process is digital, and even the final delivery to theater chains is all digital.
“In fact, Baby Driver was the first time there was no requirement for a 35mm print. For The World’s End there were still some territories which hadn’t in 2013 completely switched over to digital. But I think now, every major territory can use digital, or DCP [Digital Cinema Package].”
In spite of the complications of Baby Driver’s digital-to-film-to-digital workflow, Paul is a fan.
“I can absolutely see why, more than a kind of romantic ideal, Edgar likes to shoot on film. Because there is definitely something about that photochemical process which adds a kind of patina to the image, which is there from the moment you expose the negative, rather than capturing it digitally on the Alexa. We’re very lucky to shoot on film, and even though you get an excellent image when you shoot digitally, you can end up spending a little more time on the backend trying to give it that filmic look. Whereas film is film, and watching the 35mm images recalls the look of those fantastic movies you saw as a kid, and reminds you why you wanted to be a part of the movie world in the first place.”
This whole process of shooting Baby Driver, from choreography to cutting, was fascinating. But before I let Paul get off the phone, I couldn’t resist to ask him one last question. Something else I haven’t seen addressed in any other interviews with him.
What Drives and Defines Edgar’s Style
To prepare for this interview, I re-watched Tony Zhou’s excellent “Every Frame a Painting” episode about how to do visual comedy. I also watched the Nerdwriter’s video essay about Wright’s use of transitions in Scott Pilgrim vs. the World. Both focus on the specificity of Wright’s style and Tony specifically points out the difference he finds between American comedies and British ones. A lot of the distinction you see in Wright’s films you could attribute to the editing; particularly in the transitions in Scott Pilgrim. So I ask Paul…
“Edgar is a filmmaker who has a very distinct style? So, how much of that style is in the shooting, and how much is in the editing. Are you really the reason he has a style?” I laugh.
I half expected either a funny, self-effacing retort, or overtly comedic self-praise. What I got instead was actually rather poignant and felt totally authentic.
“It’s very fair to say actually, that a LOT of this, most of this, comes from Edgar. The reason why it looks so good is because he’s not leaving a lot of things to chance. He doesn’t just turn up on the set and go, ‘Oh, I wonder how we’re going to film this today?’ It works because he so meticulously plans these things out. He’ll spend a year or two on the animatics.
“A lot of the reviews of Baby Driver have said that the editing is great, and what Paul and Jon have done, they’ve made it all line up, and that’s true, to an extent. But of course that wouldn’t work unless you have incredible choreographers, incredible cinematographers, Edgar’s ideas in the first place. You know, he doesn’t bring us the footage and go ‘How do I put all that together and make it work on the beat?’ This has been worked out months, years, beforehand. Especially when Edgar sets himself the task of saying, ‘I want to do all of this in-camera.’
“Yes, it doesn’t just all happen magically by editing. Jon and I take our share of the credit for having it work out as smooth as it does, and problem-solving, and making it look like there weren’t any issues, but so much of that has to go to the people who are doing it there on the night. So, you know, we do our bit.”
Final Insights on Editing and the Language of Cinema
I wanted to delve deeper into this distinction of style and editing. In Tony’s video, he makes a point to mention the difference he notices between American films and British films. What he celebrated so much in the video essay about visual comedy, was Wright’s use of the camera and the visual language of film. (e.g. how people and things enter and exit frames; how shots are composed; how the camera moves; how all these things in and of themselves serve the comedy; etc.) Paul naturally had seen that video essay, and he had this to day about directors and the cinematic language.
“I’ve worked with some brilliant comedy writers and directors, who actually aren’t that interested in what the camera is doing. The camera is there to capture the comedy. And there are some shows I have worked on in the past where the director says, ‘Don’t cut so much. I have the three people in the shot, they’re providing the laughs.’ So, I know editing can help with that. But I’m a big fan of just actually having the shot there, play it long, and let them do their own natural timing in their own natural rhythm. And there is absolutely a lot to be said for that. And I learned a lot from working with these people as well, that comedy is not always just about cutting on every line. You’ve got so many talented people in front and behind the camera, sometimes it’s good just to have a shot breathe.”
If you haven’t seen Baby Driver yet, make like the titular character and speed down to your local theater and treat yourself to a cinematic delight of characters and car chases, choreography and cutting.