Behind the Seen

by Charles Koppelman

  Director of photography John Seale (behind the camera) on the set of Cold Mountain.

  August 3, 2002, Murch’s Journal

  Screening all unscreened material. Trying to catch up

  Sean still needs to find a solution to the audio file conversion problem that arose before the Final Cut Pro systems arrived from Los Angeles. This bottleneck costs him precious hours syncing up sound in time for the next day’s screening of dailies. Ramy reports to Sean that he is trying to work something out with AVTransfer, the company in Australia. Sean writes back to Ramy, pleased that DFT is putting out another call for help on sound: “Thank you, thank you. We are in this weekend, catching up and lying low. Walter continues to be a happy camper, so at least we’re doing something right.”

  Two unsolved problems remain with the Final Cut Pro workflow: 1) converting sound files from .wav format (required by the telecine in the Kodak lab) into a format that can be recognized by FCP and ProTools; and 2) Walter’s inability to get accurate results using his preferred edit-on-the-fly technique on color-corrected clips. Two longer-range problems are on the back burner: getting audio files from Final Cut Pro out to ProTools for sound editing; and getting reliable change lists to conform the 35mm picture when the film workprint is cut and recut over the course of post production.

  On the other hand, by early August, the DVD authoring feature of Final Cut has been put into operation, and copies of the dailies are made into DVDs and sent to Anthony, Sydney Pollack, and the other producers. Minghella starts building a folder containing a disc with each day’s work. This means he can review shots anytime and anywhere using his PowerBook.

  Murch works on the battle sequence.

  A few days later, as the battle sequence Murch puts together reaches 16 minutes, he suspects another long-term problem may be lurking: FCP seems to have a hard time ingesting long sequences—a concern that came up in Los Angeles but was never fully tested. It’s taking over a minute for each version of the battle sequence to load into FCP. Since Murch is working with several versions at the same time, “that means 3-4 minutes of video prep as the FCP unit is firing up,” he writes Ramy. Is this a smoldering fire that might burst into flames? Will Final Cut be able to load a five-hour assembly when the time comes, three months from now?

  On this basic function, Avid systems work on a completely opposite principle from Final Cut Pro. When Avid is loading a project, regardless of its length—an entire two-hour film, say—Avid only loads small sections of media, the idea being that Avid provides the editor with just the media he or she needs for the part of a show that is under construction. Normally an Avid editor adjusts this setting so the machine loads a single three- or four-minute chunk of digitized video and sound at a time. Within that section, operations such as edits, trims, saves, etc. happen quickly and efficiently. However, if he or she wants to jump to material that exists outside that short swatch, the Avid editor will have to wait on the machine. Final Cut Pro has no similar adjustable setting to control the amount of media the editor wants to load; it simply loads an entire sequence from beginning to end. Murch knows this is a structural problem that goes to the basic engineering of FCP that may eventually get fixed. So he asks Ramy, “Any procedural suggestions in the meantime?”

  There is no immediate solution forthcoming from DFT, but Walter remains unruffled. In fact, he flies to Paris three days later to meet Aggie and celebrate their 37th anniversary over a four-day weekend. On Murch’s return, he learns that Mihai Bogdan, the Romanian driver assigned to the edit crew, has hepatitis. The whole edit group must get gamma globulin shots, on top of the anti-rabies vaccine recommended for Romania.

  Murch was happy with Final Cut Pro’s jog wheel, which moves images backward or forward at varying speeds.

  Murch gets back in contact with Ramy. Before reminding him about the FCP load-in problem, Walter delivers an upbeat report: “FCP performed very well. I am particularly happy with the jog wheel, which is the most ‘film-like’ digital scrolling device I have experienced. I can really convince myself that I am looking at film going backward and forward at various speeds.” But the problem of how FCP handles media has just cropped up in another function. When Murch performed a routine deletion—removing an unused empty soundtrack from his Timeline, the graphic representation for a sequence—and FCP took 45 seconds to execute the command. What was a smoke alarm is now a bell-ringer. Walter knows if such a simple action takes that much time (it doesn’t even involve actual digitized video material) he might be in for some really long waits once he tries to load hours of material into Final Cut’s memory.

  Walt Shires follows up for DFT by email. He tells Murch that Final Cut Pro plays back an edited sequence by building a virtual movie (called a .moov) that is held in its memory cache, not written onto a hard drive. “You are right,” Shires writes, “the longer and more complex the edited sequence gets, the longer it will take to build the .moov. The length of time it takes to organize the .moov gets exponentially longer as the sequence gets more complex. Final Cut builds the whole movie at once so you can review any portion at any time.” One can feel Shire’s anxiety on Walter’s behalf: “There is a trade-off. Perhaps good, perhaps not. Hope your experience continues to be a good one. Walt.”

  Romanian crew member Mihai Bogdan who was promoted from driver for the edit crew to syncing up dailies.

  Murch appreciates Shire’s explanation, but he is blunt: “This is the one aspect of FCP that I have my eye on as being a potential troublemaker down the line—the [first assembly of the] film will eventually be over four hours long [over five, as it turns out]—the wait is going to become increasingly frustrating.” Murch says this issue is high on his list of “musts” for Apple to work out in subsequent FCP versions. It should be able to use buffers, like a computer printer, or an online media player, to absorb chunks of data and spool it out as needed. “Certainly,” Murch continues, “something as procedurally brainless as deleting an empty audio track should not take 45 seconds to accomplish. Otherwise, all is well. How are we moving on the change lists and OMF output fronts?”

  Aside from fretting about the capacity of his editing system Murch must attend to standard editor’s duties: informing Anthony and the producers about “negative scratches—not too serious;” providing notes to help guide the chapel-building scene which is being re-shot due to weather; and an email to producer Bill Horberg about the lead actors. “Two worry beads for me: Ada’s accent (nasal, hard to understand, very few consonants) and Inman’s hat (not cool, like Oakley’s)—it makes him skulk, if worn too low on the brow, too Hobbit-like. A girl would wonder about a guy who wore a hat like that. And a guy would wonder about a girl with a voice like that. Hmm... I have talked to Anthony about these things.”

  * * *

  Murch Email to Minghella August 14, 2002

  I am struggling a bit, just with the amount of material and keeping all the other pots on the boil, but stirring away gaily.

  * * *

  Anthony has a chat with the dialogue coach. Afterward, when he screens new dailies of Ada, Murch’s concerns about Kidman’s accent and speaking style are allayed. As for the hat Jude Law wears, Horberg writes, “Hats are Ann Roth’s domain—tread there at your own personal risk!!!”

  August 16, 2002, Murch’s Journal

  Finished the battle “work-in-progress” and sent it off to Anthony with the dailies. Let’s see what his reaction is. Screened some dailies to catch up. Need to screen two days worth each day at least to get up to speed. When will Sean be ready to give me more scenes I can cut!!

  Murch is impatient. Although he has begun cutting the battle scene, he’s ready to do more. And footage is really beginning to accumulate. Getting started two weeks late is having its effect. The equipment from DigitalFilm Tree arrived safe and sound on July 15, the first day of principal photography. But in Murch’s ideal universe it should have been in Romania in time to be plugged in, tweaked, and operating—all set to go for first day
of dailies. Sean Cullen later describes how restless Murch felt as Cullen put the system through all the necessary tests, methodically taking a piece of edited film through the entire system: “I still had to wring out the system and then wring out our workflow. This meant digitizing material, cutting a sample together on another machine (to simulate Walter’s), making tapes and outputting cut lists to check that the whole system would work in post. Walter got to see me cutting material together but he couldn’t start yet.” To allow Murch to begin editing something—the battle scene—Cullen gave him a series of one-hour QuickTime movie files. Each QuickTime movie represented one complete reel of film telecined to videotape. Even though these one-hour QuickTime files let Murch get started, it would turn out later that they were the cause of his frustration with the long load-in time he was now experiencing on Final Cut.

  Assistant Sean Cullen preparing digitized film footage for Murch to cut.

  Up in Brasov, on the set, Minghella looks at the DVD on his laptop with the battle scene Murch has edited. He sends an email:

  * * *

  Date: 8/17/02 9:14 PM

  From: Anthony Minghella

  To: Walter Murch

  Thanks, Walter

  just had some time to look at the battle sequence, for which many thanks. It has, of course, your beautiful taste and touch, and is encouraging. I hate myself, as you would expect, and my failings, and I also see how much we will have to do to get the sequence right, and how much I have ahead to make the film work between that stuff and what I’m doing now. I can’t really look at it now with a sane eye, and - naturally - mourn what isn’t there yet as much as I do what is already there, with its frailties. But I also can’t wait to stop collecting the material and sit with you in Hampstead and begin the fun.

  thanks, dear friend

  from rainy Brasov

  ant

  * * *

  Cullen’s workflow for getting shot film onto Murch’s desktop for editing.

  “Anthony sent a sad email after looking at the sketch of the battle,” Walter writes in his Journal on August 18. “I thought somewhere it would make him happy, but it just seemed to depress him more.”

  If Murch’s spirit is dampened by Minghella’s response, it doesn’t last long. The next day Sean solves the problem with loading media. On advice from DFT—“Don’t put your entire project in one FCP file”—Sean breaks the media down into bite-sized pieces for Final Cut Pro to digest.

  August 19, 2002, Murch’s Journal

  Hooray! Clips are in FCP. And when a sequence is opened up, it does so 150 times faster. An hour-long sequence opens in two seconds as opposed to four minutes. Why? Probably because the sequence is assembling from short clips of shots rather than long clips of tapes. A huge relief, that it functions so well now—that was a thundercloud bearing down on me as I thought of the problems further along in the process.

  “This was a huge turning point, in retrospect,” Murch says afterward. “It wouldn’t have been possible to keep going if it hadn’t worked out this way.” Without this workaround, Murch would be waiting longer and longer—up to a half hour or more—for his project to load into Final Cut Pro as the assembly grew in size. That kind of delay would be unacceptable in the rush to finish a major motion picture.

  The problem with large amounts of media and Final Cut Pro wasn’t an issue on previous feature films, such as The Rules of Attraction and Full Frontal, because, according to Ramy, “they got used to it.” As with many off-the-shelf features in FCP, editors simply accept the default settings, so to speak. But Murch wasn’t going to be satisfied waiting on the equipment. Sean ran with DFT’s suggestion by making separate QuickTime movies for each “flash-to-flash” take. These are discrete camera takes marked at the beginning by one or two frames of unusable bright images when the camera starts. The motor isn’t yet up to speed, so the film gets overexposed for a brief moment. A similar bright flash occurs at the end of a take when the camera is turned off and two or three frames of film are overexposed as the motor slows to a stop. By using a QuickTime movie for each flash-to-flash segment, and then linking all the adjacent QuickTime movies back together, FCP could play them seamlessly. The whole equaled its parts. Walter’s faith is rewarded after all. There was a better method for loading media; it was just a matter of discovering it.

  * * *

  Murch’s Calculations Timing Cold Mountain

  * * *

  Murch had started projecting the total amount of footage to expect on Cold Mountain one week after principal photography began. His yardstick then was the average length of each take compared to its projected length when those script pages were read aloud. At that point, take timing exceeded read-script timing by an expansion factor of 1.38. He projected the first assembly to be just over 3½ hours.

  Now, four weeks later, Murch can use a more precise tool to make his estimate—an expansion factor that uses the relationship between the timing of an edited scene and the timing of that scene when it was read aloud.

  August 20, 2002, Murch’s Journal

  Scene 12 all cut together is 144 seconds long, or exactly twice what it timed out at. So the first assembly of this film could be five and a half hours long. Read this in five months and tell me what happened.

  For Murch, this is further proof that Cold Mountain is going to have a long first assembly—longer than either English Patient or The Talented Mr. Ripley. He had hoped that after the first three weeks of filming the 12-camera action scenes of the Battle of the Crater, Cold Mountain’s expansion factor would shrink as Minghella moved into more conventional dialogue scenes. But the early indications are that this is not happening: the expansion rate stays the same.

  This means that the assembly will most likely be over five hours, and that consequently the 30 percent barrier—and its implications of open-heart surgery—will indeed have to be crossed somewhere down the line in London, during editing.

  The ultimate scheduling question is: Can he and Minghella get the film edited in time for a Christmas, 2003 release? But the high expansion rate has short-term consequences as well. Can the assistants prepare dailies for Minghella and the producers on a timely basis? Does Murch have enough time to screen and take notes on all the footage? Will he be able to prepare a first assembly by the time production wraps in mid-December? With all these demands, will Murch also be able to have a creative connection with the material so he can adequately advise Minghella about story or character issues while still in production?

  The question of keeping up with the amount of shot footage, and the inevitability of crossing the 30 percent threshold both lie on top of the fact that Murch, Cullen, and DigitalFilm Tree have just shaken out a new editing platform with several key functions that are still unproven. That should certainly be enough. But there are other, more unwelcome surprises.

  August 27, 2002, Murch’s Journal

  Driver crashed on way back from Brashov and the negative scattered all over the road. All ok, though—the cans didn’t open up.

  To be followed by agreeable arrivals.

  August 30, 2002, Murch’s Journal

  Aggie arrives today she is just getting on the plane as I write this!

  Aggie arrived! At 6.30—great to see her—she brought gifts for everyone (tea and Marmite) and everyone loved to see her. Especially me.

  At the end of August, the Cold Mountain production moves to North Carolina for three weeks to shoot various scenes there: Inman’s walk from sea to mountains; the hospital and the peanut seller; Veaseytown; the ferry crossing; and Veasey’s discovery of the saw. Murch and the edit crew stay in Romania, so this is a chance to get caught up with logging, digitizing, and other housekeeping duties, since it will be two weeks before dailies footage from the U.S. arrives in Bucharest.

  * * *

  Muriel Ann Murch, aka Aggie

  Aggie is amazing. She has the most amazing influence on me. She makes me feel eight years old, but a happy eight. Most of the time I feel eight y
ears old but a sad eight.

  —Anthony Minghella

  * * *

  Minghella sends Murch an email report from the U.S.:

  * * *

  Date: 9/2/02 4:04 PM

  From: Anthony Minghella

  To: Walter Murch

  Subject: Charleston

  ...where, apparently, the drought they’re suffering is about to be solved by a huge tropical storm! I am plagued.

  Happy to get some dailies on DVD although they don’t all play well. Kiss interrupted by skipping disk.

  [Cold Mountain author Charles] Frazier saw some stuff and seemed cheered.

  Should I try and get a close-up of Butcher lying dead on the ground in the night raid? He lives in Charleston.

  Having a free day here for Labor Day after an interminable journey immediately followed by a day’s scouting. Relentless.

  Very prone to depression on this movie. Which was also the case shooting The English Patient. What is it? Why is it? Just tired, I suppose. But trying to find a way to take pleasure from the challenge of collecting the material and not just to be ground down.

  Love to you. Very glad you’re out there helping

  xa

  * * *

  * * *