Birdman

Birdman-PosterIt’s rare, but exhilarating, when you watch a movie with a unique take on film’s visual language, without the crutch of extensive computer generated imagery. That’s precisely the beauty of Birdman or (The Unexpected Virtue of Ignorance). The film is directed and co-written by Alejandro González Iñárritu (Biutiful, Babel, 21 Grams) and features a dynamic ensemble cast dominated by Michael Keaton’s lead performance as Riggan Thomson. While most films are constructed of intercutting master shots, two-shots and singles, Birdman is designed to look like a continuous, single take. While this has been done before in films, approximately 100 minutes out of the two-hour movie appear as a completely seamless composite of lengthy Steadicam and hand-held takes.

Riggan Thomson (Keaton) is a movie star who rode to fame as the comic book super hero Birdman; but, it’s a role that he walked away from. Searching for contemporary relevance, Riggan has decided to mount a Broadway play, based on the real-life Raymond Carter short story, What We Talk About When We Talk About Love. The film takes place entirely at the historic St. James Theater near Times Square and the surrounding area in New York. Principal photography occurred over a 30-day period, both at the real theater and Times Square, as well as at Kaufman Astoria Studios. The soundstage sets were for the backstage and dressing room portions of the theater. Throughout the film, Riggan struggles with the challenges of getting the play to opening day and dealing with his fellow cast members, but more notably confronting his super ego Birdman, seen in person and heard in voice-over. This is, of course, playing out in Riggan’s imagination. The film, like the play within the film, wrestles with the overall theme of the confusion between love and affection.

Bringing this ambitious vision to life fell heavily upon the skills of the director of photography and the editors. Emmanuel Lubezki, known as Chivo, served as DoP. He won the 2014 Cinematography Oscar for Gravity, a film that was also heralded for its long, seemingly continuous shots. Stephen Mirrione (The Monuments Men, Ocean’s Thirteen, Traffic) and Douglas Crise (Arbitrage, Deception, Babel) teamed up for the edit. Both editors had worked together before, as well as with the director. Mirrione started during the rehearsal process. At the time of production, Crise handled the editing in New York, while Mirrione, who was back LA at this time, was getting dailies and checking in on the cut as well as sending scenes back and forth with changes every day.

It starts with preparation

Stephen Mirrione explains, “When I first saw what they wanted to do, I was a bit skeptical that it could be pulled off successfully. Just one scene that didn’t work would ruin the whole film. Everything really had to align. One of the things that was considered, was to tape and edit all of the rehearsals. This was done about two months before the principal photography was set to start. The rehearsals were edited together, which allowed Alejandro to adjust the script, pacing and performances. We could see what would work and what wouldn’t. Before cameras even rolled, we had an assembly made up of the rehearsal footage and some of the table read. So, together with Alejandro, we could begin to gauge what the film would look and sound like, where a conversation was redundant, where the moves would be. It was like a pre-vis that you might create for a large-scale CGI or animated feature.”

Once production started in New York, Douglas Crise picked up the edit. Typically, the cast and crew would rehearse the first half of the day and then tape during the second half. ARRI ALEXA cameras were used. The media was sent to Technicolor, who would turn around color corrected Avid DNxHD36 dailies for the next day. The team of editors and assistants used Avid Media Composer systems. According to Crise, “I would check the previous day’s scenes and experiment to see how the edit points would ‘join’ together. You are having to make choices based on performance, but also how the camera work would edit together. Alejandro would have to commit to certain editorial decisions, because those choices would dictate where the shot would pick up on the next day. Stephen would check in on the progress during this period and then he picked up once the cut shifted to visual effects.”

Naturally the editing challenge was to make the scenes flow seamlessly in both a figurative and literal sense. “The big difference with this film was that we didn’t have the conventional places where one scene started and another ended. Every scene walks into the next one. Alejandro described it as going down a hill and not stopping. There wasn’t really a transition. The characters just keep moving on,” Crise says.

“I think we really anticipated a lot of the potential pitfalls and really prepared, but what we didn’t plan on were all the speed changes,” Mirrione adds. “At certain points, when the scene was not popping for us, if the tempo was a little off, we could actually dial up the pace or slow it down as need be without it being perceptible to the audience and that made a big difference.”

Score and syncopation

To help drive pace, much of the track uses a drum score composed and performed by Mexican drummer Antonio Sanchez. In some scenes within the film, the camera also pans over to a drummer with a kit who just happens to be playing in an alley or even in a backstage hallway. Sanchez and Iñárritu went into a studio and recorded sixty improvised tracks based on the emotions that the film needed. Mirrione says, “Alejandro would explain the scene to the drummer in the studio and then he’d create it.” Crise continues, “Alejandro had all these drum recordings and he told me to pick six of my favorites. We cut those together so that he could have a track that the drummer could mimic when they shot that scene. He had the idea for the soundtrack from the very beginning and we had those samples cut in from the start, too.”

“And then Martín [Hernández, supervising sound editor] took it to another level. Once there was an first pass at the movie, with a lot of those drum tracks laid in as an outline, he spent a lot of time working with Alejandro, to strip layers away, add some in, trying a lot of different beats. Obviously, in every movie, music will have an impact on point of view and mood and tone. But with this, I think it was especially important, because the rhythm is so tied to the camera and you can’t make those kinds of cadence adjustments with as much flexibility as you can with cuts. We had to lean on the music a little more than normal at times, to push back or pull forward,” Mirrione says.

The invisible art

The technique of this seamless sequence of scenes really allows you to get into the head of Riggan more so than other films, but the editors are reserved in discussing the actual editing technique. Mirrione explains, “Editing is often called the ‘invisible art’. We shape scenes and performances on every film. There has been a lot of speculation over the internet about the exact number and length of shots. I can tell you it’s more than most people would guess. But we don’t want that to be the focus of the discussion. The process is still the same of affecting performance and pace. It’s just that the dynamic has been shifted, because much of the effort was front-loaded in the early days. Unlike other films, where the editing phase after the production is completed, focuses on shaping the story – on Birdman it was about fine-tuning.”

Crise continues, “Working on this film was a different process and a different way to come up with new ideas. It’s also important to know that most of the film was shot practically. Michael [Keaton] really is running through Times Square in his underwear. The shots are not comped together with green screen actors against CGI buildings.” There are quite a lot of visual effects used to enhance and augment the transitions from one shot to the next to make these seamless. On the other hand, when Riggan’s Birdman delusions come to life on screen, we also see more recognizable visual effects, such as a brief helicopter and creature battle playing out over the streets of New York.

Winking at the audience

The film is written as a black comedy with quite a few insider references. Clearly, the casting of Michael Keaton provides allusion to his real experiences in starting the Batman film franchise and in many ways the whole super hero film genre. However, there was also a conscious effort during rehearsals and tapings to adjust the dialogue in ways that kept these references as current as possible. Crise adds, “Ironically, in the scenes on the rooftop there was a billboard in the background behind Emma Stone and Edward Norton, with a reference to Tom Hanks. We felt that audiences would believe that we created it on purpose, when if fact it was a real billboard. It was changed in post, just to keep from appearing to be an insider reference that was too obvious.”

The considerations mandated during the edit by a seamless film presented other challenges, too. For example, simple concerns, like where to structure reel breaks and how to hand off shots for visual effects. Mirrione points out, “Simple tasks such as sending out shots for VFX, color correction, or even making changes for international distribution requirements were complicated by the fact that once we finished, there weren’t individual ‘shots’ to work with – just one long never ending strand.  It meant inventing new techniques along the way.  Steven Scott, the colorist, worked with Chivo at Technicolor LA to perfect all the color and lighting and had to track all of these changes across the entire span of the movie.  The same way we found places to hide stitches between shots, they had to hide these color transitions which are normally changed at the point of a cut from one shot to the next.”

In total, the film was in production and post for about a year, starting with rehearsals in early 2013. Birdman was mixed and completed by mid-February 2014. While it posed a technical and artistic challenge from the start, everything amazingly fell into place, highlighted by perfect choreography of cast, crew and the post production team. It will be interesting to see how Birdman fares during awards season, because it succeeds on so many levels.

Originally written for Digital Video magazine / CreativePlanetNetwork

©2014, 2015 Oliver Peters

Stocking Stuffers 2014

df_stuff14_1_smAs we head toward the end of the year, it’s time to look again at a few items you can use to spruce up your edit bay.

Let’s start at the computer. The “tube” Mac Pro has been out for nearly a year, but many will still be trying to get the most life out of their existing Mac Pro “tower”. I wrote about this awhile back, so this is a bit of a recap. More RAM, an internal SSD and an upgraded GPU card are the best starting points. OWC and Crucial are your best choices for RAM and solid state drives. If you want to bump up your GPU, then the Sapphire 7950 (Note: I have run into issues with some of these cards, where the spacer screws are too tall, requiring you to install the card in slot 2) and/or Nvidia GTX 680 Mac Edition cards are popular choices. However, these will only give you an incremental boost if you’ve already been running an ATI 5870 or Nvidia Quadro 4000 display card. df_stuff14_2_smIf you have the dough and want some solid horsepower, then go for the Nvidia Quadro K5000 card for the Mac. To expand your audio monitoring, look at Mackie mixers, KRK speakers and the PreSonus Audiobox USB interface. Naturally there are many video monitor options, but assuming you have an AJA or Blackmagic Design interface, FSI would be my choice. HP Dreamcolor is also a good option when connecting directly to the computer.

The video plug-in market is prolific, with plenty of packages and/or individual filters from FxFactory, Boris, GenArts, FCP Effects, Crumplepop, Red Giant and others. I like the Universe package from df_stuff14_3_smRed Giant, because it supports FCP X, Motion, Premiere Pro and After Effects. Red Giant continues to expand the package, including some very nice new premium effects. If you are a Media Composer user, then you might want to look into the upgrade from Avid FX to Boris Red. Naturally, you can’t go wrong with FxFactory, especially if you use FCP X. There’s a wide range of options with the ability to purchase single filters – all centrally managed through the FxFactory application.

df_stuff14_4_smFor audio, the go-to filter companies are iZotope, Waves and Focusrite to name a few. iZotope released some nice tools in its RX4 package – a state-of-the-art repair and restoration suite. If you just want a suite of EQ and compression tools, then Nectar Elements or Nectar 2 are the best all-in-one collections of audio filters. While most editors do their audio editing/mastering within their NLE, some need a bit more. Along with a 2.0 bump for Sound Forge Pro Mac, Sony Creative Software also released a standard version of Sound Forge through the Mac App Store.

df_stuff14_5_smIn the color correction world, there’s been a lot of development in film emulation look-up tables (LUTs). These can be used in most NLEs and grading applications. If that’s for you, check out ImpulZ and Osiris from Color Grading Central (LUT Utility required with FCP X), Koji Color or the new SpeedLooks 4 (from LookLabs). Each package offers a selection of Fuji and Kodak emulations, as well as other stylized looks. These packages feature LUT files in the .cube and/or .look (Adobe) LUT file formats and, thus, are compatible with most applications. If you want film emulation that also includes 3-way grading tools and adjustable film grain, your best choice is FilmConvert 2.0.

df_stuff14_6_smAnother category that is expanding covers the range of tools used to prep media from the camera prior to the edit. This had been something only for DITs and on-set “data wranglers”, but many videographers are increasingly using such tools on everyday productions. These now offer on-set features that benefit all file-based recordings. Pomfort Silverstack, ShotPut Pro, Redcine-X Pro and Adobe Prelude have been joined by new tools. To start, there’s Offload and EditReady, which are two very specific tools. Offload simply copies and verifies camera-card media to two target drives. EditReady is a simple drag-and-drop batch convertor to transcode media files. These join QtChange (a utility to batch-add timecode and reel IDs to media files) and Better Rename (a Finder renaming utility) in my book, as the best single-purpose production applications.

df_stuff14_7_smIf you want more in one tool, then there’s Bulletproof, which has now been joined in the market by Sony Creative Software’s Catalyst Browse and Prepare. Bulletproof features media offload, organization, color correction and transcoding. I like it, but my only beef is that it doesn’t properly handle timecode data, when present. Catalyst Browse is free and similar to Canon’s camera utility. It’s designed to read and work with media from any Sony camera. Catalyst Prepare is the paid version with an expanded feature set. It supports media from other camera manufacturers, including Canon and GoPro.

df_stuff14_8_smFinally, many folks are looking for alternative to Adobe Photoshop. I’m a fan of Pixelmator, but this has been joined by Pixlr and Mischief. All three are available from the Mac App Store. Pixlr is free, but can be expanded through subscription. In its basic form, Pixlr is a stylizing application that is like a very, very “lite” version of Photoshop; however, it includes some very nice image processing filters. Mischief is a drawing application designed to work with drawing tablets, although a mouse will work, too.

©2014 Oliver Peters

DaVinci Resolve 11

df_resolve11_1_sm

With the release of DaVinci Resolve 11, Blackmagic Design has firmly moved into the ranks of nonlinear editing. In addition to a redesigned logo and splash screen, Resolve 11 sports more editorial tools than ever before. Now for the first time it is worthy for consideration as your NLE of choice. I have covered previous releases of Resolve, so I’ll only briefly touch on color correction in this article.

As before, DaVinci Resolve 11 comes in four versions: Resolve Lite (free), Resolve Software ($995), Resolve (with the control surface for $29,995) and the Linux configuration. Both free and paid versions support a variety of third-party control surfaces, with the most popular being the Avid Artist Color and the Tangent Devices Element panels. Resolve Lite supports output up to UltraHD (3840 x 2160). It includes most of the features of the paid software, except collaboration, stereo 3D and noise reduction. Although you can operate Resolve without any third-party i/o hardware, if you want external monitoring or output to tape, you’ll need to purchase one of Blackmagic Design’s PCIe capture cards or Thunderbolt i/o devices.

Color Match

df_resolve11_2_smThe interface is divided into four modules: media, edit, color and deliver. All color correction occurs in the color module. Here you’ll find a wealth of grading tools, including camera raw settings, color wheels, primary sliders and more. Color Match is a new correction tool. If you included a color chart when you shot your footage, Resolve can use the image of that chart to set an automatic correction for the color balance of the scene.

Color Match features three template settings for charts, including X-Rite ColorChecker, Datacolor SpyderCheckr and DSC Labs OneShot. If you used one of these charts and it’s in your footage, then select the appropriate set of color swatches in the Color Match menu. Next, select the Color Chart grid from the viewer tools, which opens an overlay for that chart. Corner-pin the overlay so that the grid lines up over the color swatches in the image and hit the Match button. Resolve will instantly adjust its curves to correct the color balance of the shot, so that the chart in the image matches the template for that chart in Resolve. Now you can copy this grade and apply it to the rest of the shots within that same set-up.

Although this isn’t a one-shot fix, it’s intended to give you a good starting point for your grade. While this feature demos really well and is certainly a whizz-bang attention-getter, it has the most value for novice users or for DITs who need to get a quick grade for dailies while on location.

Editing

df_resolve11_3_smThe biggest spark of interest I’ve seen for Resolve 11 is due to the editing tools. As an NLE, it’s somewhat of a mash-up between Final Cut Pro 7 and Final Cut Pro X. It copies a lot of X’s design aesthetic and even some features, like clip skimming in the media bins; yet, it is clearly track-based. For editors who like a lot of FCP X, but are put off by Apple’s trackless, magnetic timeline, Resolve 11 becomes a very tantalizing, cross-platform alternative.

Resolve 11’s edit module most closely aligns with Final Cut Pro 7, although there is no multi-cam feature, yet. The keyboard commands mirror the FCP 7 set, as do menu options and much of the working style. One big improvement is a very advanced trim mode, which offers good asymmetrical trimming. If you start a project from the beginning in Resolve 11, you can easily import media, organize clips into bins/folders, add logging information and, in general, do all of the nuts-and-bolts things you do in every editing application.

The interface uses a modal design and supports dual and single monitor configurations. Although there are numerous panels and windows that can be opened as needed, the general layout is fixed. Certain functions are restricted to the edit module and others to the color module. For example, transition effects, titles and generators would be added and adjusted in the edit module, while working with a standard timeline. Color correction and other image effects are reserved for the color module, which uses a node-based hierarchy. Resizing and repositioning can be done in either.

Resolve includes an inspector pane on the right side of the timeline viewer that is much like that of FCP X. Here you’ll find composite, transform, cropping, retiming and scaling controls. If you select a transition, then its adjustment controls appear in the inspector panel. Resolve supports the OpenFX video plug-in architecture. Third-party transitions will show up in the edit mode’s OpenFX library, while the filters only show up when you are in the color mode. Like FCP X, inspector controls are limited to sliders, color pickers and numerical entry, with no allowance for custom third-party plug-in interfaces.

My biggest beef is performance. Resolve 11 is optimized to pass the highest quality images through its pipeline, which seems to impede real-time playback, even with ungraded footage. In other NLEs, hitting play or the space bar brings you to full-speed, real-time playback in a fraction of a second. In Resolve it takes a few seconds, which is clearly evident in its dropped-frame indicator. Even with proper real-time playback, video motion does not look as smooth and fluid in the viewer as I would expect. There are a number of factors that affect this, including drive performance (high-performance storage is good), GPU performance (one or more high-end cards are desirable) and age of the machine (a new top-of-the-line system is ideal). Resolve is also not as gracious with a wide range of native media types as some of the other NLEs.

df_resolve11_4_smColor grades will affect performance. What if you start grading in the color mode and bounce back into the edit mode? This has the same impact on the computer as applying several filters in a traditional NLE. Add a stack of effects on most NLEs and playback performance through those clips is often terrible until you render. To mitigate this issue, Resolve includes smart caching, which is a similar sort of background render as that of FCP X. The software renders clips with a grade or an effect applied anytime the machine is idle.

Audio in Resolve 11 is still in the very early stages. There is no audio plug-in architecture. Hopefully Blackmagic will add AU and/or VST support down the road. Having multiple audio tracks also hurts system performance. Complex audio in the timeline quickly choked my system. Having even a few tracks caused the audio to drop out during playback. Resolve employs a similar track design to Adobe Premiere Pro. This means adaptive tracks, where a single timeline track can contain one mono channel, two stereo channels or multiple surround channels. This is an interesting design, but it seems to impact round-tripping between other applications. For example, I’ve exported multi-channel timelines via XML. In this process, when I brought that timeline into FCP 7 or Premiere Pro, these tracks only showed up as mono tracks with one channel of audio.

Roundtrips

df_resolve11_6_smWhere Resolve 11 really shines is in its roundtrip capabilities. It can take media and edit list formats from a range of systems, then let you process the media and finally output a new set of media files and corresponding lists. EDL, AAF, XML and FCPXML formats are supported, making Resolve 11 one of the better cross-application conversion tools. For instance, you can edit in FCP X, conform and grade in Resolve 11 and then output that in a compatible format to finish in the same or different NLE, such as Media Composer, Smoke, Premiere Pro, etc. Of course, with Resolve 11, you could simply finish in Resolve and output final deliverables from right within the application. That’s clearly the design goal Blackmagic had in mind.

Personally, I still prefer to use the roundtrip method, but there are a few wrinkles in this process. I have already mentioned audio issues. Another is resizing, such as FCP X’s “spatial conform” and Premiere Pro’s “scale to frame size”. These are automatic timeline functions to fit oversized images into smaller timeline frames, such as putting 4K media into a 1080 timeline. This feature automatically down-scales the source image so that either horizontal or vertical dimensions match. Unfortunately some of this information gets lost in the translation between applications.

df_resolve11_8_smI recently ran into this on two jobs with 4K RED media and Resolve 11. The first was a project cut in FCP X. The roundtrip went fine, but when the newly rendered 1080 media was back in FCP X, the application still thought it needed to enable spatial conform, which had been used in the offline edit. Disabling spatial conform caused FCP X to blow up the 1080 media 200%. The simple fix was just to leave spatial conform on and let FCP X render this media on export. There were no visible issues that I could detect.

The second was a music video project that the director had cut on Premiere Pro CC2014. There was extensive reframing and repositioning throughout. Importing this timeline into Resolve 11 was a complete disaster and would have meant rebuilding all of this work to reframe images. Ultimately I opted to use SpeedGrade CC2014 on this particular job, since it correctly translated the Premiere Pro timeline via Adobe’s Direct Link feature.

As a general rule, I would recommend that if you know you are going outside of the application, do not use any of these automatic resizing tools in the offline NLE. Instead, manually set the scale and position values, because Resolve does an excellent job of interpreting these parameters when set during the offline edit.

OpenFX

df_resolve11_5_smBlackmagic added the OpenFX architecture with Resolve 10, but now that Resolve 11 is out, new developers are joining the party. On my test system I installed both the FilmConvert 2.0 plug-in and the Boris Continuum Complete 9 package. The filters are accessed in the color modules and are applied to nodes, just like other grading functions. Although other host versions of the FilmConvert filter include color wheels within the filter’s control panel, they are excluded in the OpenFX version. You do get the camera and film emulsion presets. This is my favorite film emulation and grain plug-in and it makes a suitable complement to Resolve.

Boris FX’s BCC 9 for Resolve includes most of the same filters as for other hosts, including the new FX Browser. You can launch it from inside the Resolve interface, but when I tried to use it, the browser crashed the application. I’m running the public beta of 11.1, so that could be part of it. Otherwise, the filters themselves worked fine. So, if you need to add a glow, cartoon effect or spray paint noise to a shot, you can do so from inside Resolve with BCC 9.

OpenFX filters installed for other applications also show up in Resolve. I discovered this during my review of the HP Z1G2 workstation. Sony Vegas Pro 13 was installed, which also uses OpenFX. The NewBlueFX filters that were installed for Vegas also showed up in Resolve 11 on that machine.

A key point to remember it to apply OpenFX filters in a separate node. If you need to change the filter, simply delete the node and create a new one for a different filter. That way you won’t lose any of the correction applied to the clip.

Collaboration

df_resolve11_7_smResolve 11 enables collaboration among multiple users on the same project. This requires a paid version of Resolve 11 for each collaborator, a network and a shared DaVinci Resolve database. To test this feature, I enlisted the help of colorist and trainer Patrick Inhofer (Tao of Color, Mixing Light). Patrick set up a simple ethernet network between a Mac Pro and a MacBook Pro, each running a paid version of Resolve 11. You have to set up a shared project and open both Resolve seats in the collaboration mode. Once both systems are open with the same project, then it is possible to work interactively.

This is not like two or more Avid Media Composers running in a Unity-style sharing configuration. Rather, this approach is intended for an editor and a colorist to be able to simultaneously work on one timeline at the same time. One person is the “owner” of the project, while anyone else is a “collaborator”. In this model, the “owner” has control of the editing timeline and the “collaborator” is the colorist working in the color module. You could also have a third collaborator logging metadata for clips.

df_resolve11_9_smIn the collaboration mode, a bell-shaped alert icon is added to the lower left corner of the interface. Whenever the colorist adds or changes a correction on one or more clips and publishes his changes, the editor receives an alert to update the clips. When the update is made, the colorist’s changes become visible on the clips in the editor’s timeline. If the editor makes editorial changes to the timeline, such as trimming, adding or deleting clips, then he or she must save the project. Once saved, the colorist can reload the project to see these updates.

As long as you follow these procedures, things work well; however, in our tests, when we went the other direction, updates didn’t happen correctly. For example, color changes made by the editor or timeline edits made by the colorist, did not show up as expected on the other person’s system. Collaboration worked well, once we both got the hang of it, but the feature does feel like a 1.0 version. Updating changes worked, but you can also reject a change by choosing “revert”. This is supposed to take the clip back to the previous grade. Instead, it dropped the grade entirely and went back to an un-corrected version of the clip with all nodes removed.

DaVinci Resolve 11 is a powerful new version of this best-in-class color grading application. Although you might not edit a project from start-to-finish in Resolve, you certainly could. For now, Blackmagic Design is positioning Resolve as an NLE designed for finishing. Edit your creative cut in Media Composer, Final Cut Pro or Premiere Pro – mix in Logic Pro X, Pro Tools or Audition – and then bring them all together in Resolve 11. As we all know, clients like to tweak the cut until the very end. Now the grading environment can enjoy more interactivity than ever before.

Originally written for Digital Video magazine / CreativePlanetNetwork.

©2014 Oliver Peters

Gone Girl

df_gg_4David Fincher is back with another dark tale of modern life, Gone Girl – the film adaptation of Gillian Flynn’s 2012 novel. Flynn also penned the screenplay.  It is the story of Nick and Amy Dunne (Ben Affleck and Rosamund Pike) – writers who have been hit by the latest downturn in the economy and are living in America’s heartland. Except that Amy is now mysteriously missing under suspicious circumstances. The story is told from each of their subjective points of view. Nick’s angle is revealed through present events, while Amy’s story is told through her diary in a series of flashbacks. Through these we learn that theirs is less than the ideal marriage we see from the outside. But whose story tells the truth?

To pull the film together, Fincher turned to his trusted team of professionals including director of photography Jeff Cronenweth, editor Kirk Baxter and post production supervisor Peter Mavromates. Like Fincher’s previous films, Gone Girl has blazed new digital workflows and pushed new boundaries. It is the first major feature to use the RED EPIC Dragon camera, racking up 500 hours of raw footage. That’s the equivalent of 2,000,000 feet of 35mm film. Much of the post, including many of the visual effects, were handled in-house.

df_gg_1Kirk Baxter co-edited David Fincher’s The Curious Case of Benjamin Button, The Social Network and The Girl with the Dragon Tattoo with Angus Wall – films that earned the duo two best editing Oscars. Gone Girl was a solo effort for Baxter, who had also cut the first two episodes of House of Cards for Fincher. This film now becomes the first major feature to have been edited using Adobe Premiere Pro CC. Industry insiders consider this Adobe’s Cold Mountain moment. That refers to when Walter Murch used an early version of Apple Final Cut Pro to edit the film Cold Mountain, instantly raising the application’s awareness among the editing community as a viable tool for long-form post production. Now it’s Adobe’s turn.

In my conversation with Kirk Baxter, he revealed, “In between features, I edit commercials, like many other film editors. I had been cutting with Premiere Pro for about ten months before David invited me to edit Gone Girl. The production company made the decision to use Premiere Pro, because of its integration with After Effects, which was used extensively on the previous films. The Adobe suite works well for their goal to bring as much of the post in-house as possible. So, I was very comfortable with Premiere Pro when we started this film.”

It all starts with dailies

df_gg_3Tyler Nelson, assistant editor, explained the workflow, “The RED EPIC Dragon cameras shot 6K frames (6144 x 3072), but the shots were all framed for a 5K center extraction (5120 x 2133). This overshoot allowed reframing and stabilization. The .r3d files from the camera cards were ingested into a FotoKem nextLAB unit, which was used to transcode edit media, viewing dailies, archive the media to LTO data tape and transfer to shuttle drives. For offline editing, we created down-sampled ProRes 422 (LT) QuickTime media, sized at 2304 x 1152, which corresponded to the full 6K frame. The Premiere Pro sequences were set to 1920 x 800 for a 2.40:1 aspect. This size corresponded to the same 5K center extraction within the 6K camera files. By editing with the larger ProRes files inside of this timeline space, Kirk was only viewing the center extraction, but had the same relative overshoot area to enable easy repositioning in all four directions. In addition, we also uploaded dailies to the PIX system for everyone to review footage while on location. PIX also lets you include metadata for each shot, including lens choice and camera settings, such as color temperature and exposure index.”

Kirk Baxter has a very specific way that he likes to tackle dailies. He said, “I typically start in reverse order. David tends to hone in on the performance with each successive take until he feels he’s got it. He’s not like other directors that may ask for completely different deliveries from the actors with each take. With David, the last take might not be the best, but it’s the best starting point from which to judge the other takes. Once I go through a master shot, I’ll cut it up at the points where I feel the edits will be made. Then I’ll have the assistants repeat these edit points on all takes and string out the line readings back-to-back, so that the auditioning process is more accurate. David is very gifted at blocking and staging, so it’s rare that you don’t use an angle that was shot for a scene. I’ll then go through this sequence and lift my selected takes for each line reading up to a higher track on the timeline. My assistants take the selects and assemble a sequence of all the angles in scene order. Once it’s hyper-organized, I’ll send it to David via PIX and get his feedback. After that, I’ll cut the scene. David stays in close contact with me as he’s shooting. He wants to see a scene cut together before he strikes a set or releases an actor.”

Telling the story

df_gg_5The director’s cut is often where the story gets changed from what works on paper to what makes a better film. Baxter elaborated, “When David starts a film, the script has been thoroughly vetted, so typically there isn’t a lot of radical story re-arrangement in the cutting room. As editors, we got a lot of credit for the style of intercutting used in The Social Network, but truthfully that was largely in the script. The dialogue was tight and very integral to the flow, so we really couldn’t deviate a lot. I’ve always found the assembly the toughest part, due to the volume and the pressure of the ticking clock. Trying to stay on pace with the shoot involves some long days. The shooting schedule was 106 days and I had my first cut ready about two weeks after the production wrapped. A director gets around ten weeks for a director’s cut and with some directors, you are almost starting from scratch once the director arrives. With David, most of that ten week period involves adding finesse and polish, because we have done so much of the workload during the shoot.”

df_gg_9He continued, “The first act of Gone Girl uses a lot of flashbacks to tell Amy’s side of the story and with these, we deviated a touch from the script. We dropped a couple of scenes to help speed things along and reduced the back and forth of the two timelines by grouping flashbacks together, so that we didn’t keep interrupting the present day; but, it’s mostly executed as scripted. There was one scene towards the end that I didn’t feel was in the right place. I kept trying to move it, without success. I ended up taking another pass at the cut of the scene. Once we had the emotion right in the cut, the scene felt like it was in the right place, which is where it was written to be.”

“The hardest scenes to cut are the emotional scenes, because David simplifies the shooting. You can’t hide in dynamic motion. More complex scenes are actually easier to cut and certainly quite fun. About an hour into the film is the ‘cool girls’ scene, which rapidly answers lots of question marks that come before it. The scene runs about eight minutes long and is made up of about 200 set-ups. It’s a visual feast that should be hard to put together, but was actually dessert from start to finish, because David thought it through and supplied all the exact pieces to the puzzle.”

Music that builds tension

df_gg_6Composers Trent Reznor and Atticus Ross of Nine Inch Nails fame are another set of Fincher regulars. Reznor and Ross have typically supplied Baxter with an album of preliminary themes scored with key scenes in mind. These are used in the edit and then later enhanced by the composers with the final score at the time of the mix. Baxter explained, “On Gone Girl we received their music a bit later than usual, because they were touring at the time. When it did arrive, though, it was fabulous. Trent and Atticus are very good at nailing the feeling of a film like this. You start with a piece of music that has a vibe of ‘this is a safe, loving neighborhood’ and throughout three minutes it sours to something darker, which really works.”

“The final mix is usually the first time I can relax. We mixed at Skywalker Sound and that was the first chance I really had to enjoy the film, because now I was seeing it with all the right sound design and music added. This allows me to get swallowed up in the story and see beyond my role.”

Visual effects

df_gg_7The key factor to using Premiere Pro CC was its integration with After Effects CC via Adobe’s Dynamic Link feature. Kirk Baxter explained how he uses this feature, “Gone Girl doesn’t seem like a heavy visual effects film, but there are quite a lot of invisible effects. First of all, I tend to do a lot of invisible split screens. In a two-shot, I’ll often use a different performance for each actor. Roughly one-third of the timeline contains such shots. About two-thirds of the timeline has been stabilized or reframed. Normally, this type of in-house effects work is handled by the assistants who are using After Effects. Those shots are replaced in my sequence with an After Effects composition. As they make changes, my timeline is updated.”

“There are other types of visual effects, as well. David will take exteriors and do sky replacements, add flares, signage, trees, snow, breath, etc. The shot of Amy sinking in the water, which has been used in the trailers, is an effects composite. That’s better than trying to do multiple takes with the real actress by drowning her in cold water. Her hair and the water elements were created by Digital Domain. This is also a story about the media frenzy that grows around the mystery, which meant a lot of TV and computer screen comps. That content is as critical in the timing of a scene as the actors who are interacting with it.”

Tyler Nelson added his take on this, “A total of four assistants worked with Kirk on these in-house effects. We were using the same ProRes editing files to create the composites. In order to keep the system performance high, we would render these composites for Kirk’s timeline, instead of using unrendered After Effects composites. Once a shot was finalized, then we would go back to the 6K .r3d files and create the final composite at full resolution. The beauty of doing this all internally is that you have a team of people who really care about the quality of the project as much as everyone else. Plus the entire process becomes that much more interactive. We pushed each other to make everything as good as it could possibly be.”

Optimization and finishing

df_gg_2A custom pipeline was established to make the process efficient. This was spearheaded by post production consultant Jeff Brue, CTO of Open Drives. The front end storage for all active editorial files was a 36TB RAID-protected storage network built with SSDs. A second RAID built with standard HDDs was used for the .r3d camera files and visual effects elements. The hardware included a mix of HP and Apple workstations running with NVIDIA K6000 or K5200 GPU cards. Use of the NVIDIA cards was critical to permit as much real-time performance as possible doing the edit. GPU performance was also a key factor in the de-Bayering of .r3d files, since the team didn’t use any of the RED Rocket accelerator cards in their pipeline. The Macs were primarily used for the offline edit, while the PCs tackled the visual effects and media processing tasks.

In order to keep the Premiere Pro projects manageable, the team broke down the film into eight reels with a separate project file per reel. Each project contained roughly 1,500 to 2,000 files. In addition to Dynamic Linking of After Effects compositions, most of the clips were multi-camera clips, as Fincher typically shoots scenes with two or more cameras for simultaneous coverage. This massive amount of media could have potentially been a huge stumbling block, but Brue worked closely with Adobe to optimize system performance over the life of the project. For example, project load times dropped from about six to eight minutes at the start down to 90 seconds at best towards the end.

The final conform and color grading was handled by Light Iron on their Quantel Pablo Rio system run by colorist Ian Vertovec. The Rio was also configured with NVIDIA Tesla cards to facilitate this 6K pipeline. Nelson explained, “In order to track everything I used a custom Filemaker Pro database as the codebook for the film. This contained all the attributes for each and every shot. By using an EDL in conjunction with the codebook, it was possible to access any shot from the server. Since we were doing a lot of the effects in-house, we essentially ‘pre-conformed’ the reels and then turned those elements over to Light Iron for the final conform. All shots were sent over as 6K DPX frames, which were cropped to 5K during the DI in the Pablo. We also handled the color management of the RED files. Production shot these with the camera color metadata set to RedColor3, RedGamma3 and an exposure index of 800. That’s what we offlined with. These were then switched to RedLogFilm gamma when the DPX files were rendered for Light Iron. If, during the grade, it was decided that one of the raw settings needed to be adjusted for a few shots, then we would change the color settings and re-render a new version for them.” The final mastering was in 4K for theatrical distribution.

df_gg_8As with his previous films, director David Fincher has not only told a great story in Gone Girl, but set new standards in digital post production workflows. Seeking to retain creative control without breaking the bank, Fincher has pushed to handle as many services in-house as possible. His team has made effective use of After Effects for some time now, but the new Creative Cloud tools with Premiere Pro CC as the hub, bring the power of this suite to the forefront. Fortunately, team Fincher has been very eager to work with Adobe on product advances, many of which are evident in the new application versions previewed by Adobe at IBC in Amsterdam. With a film as complex as Gone Girl, it’s clear that Adobe Premiere Pro CC is ready for the big leagues.

Kirk Baxter closed our conversation with these final thoughts about the experience. He said, “It was a joy from start to finish making this film with David. Both he and Cean [Chaffin, producer and David Fincher’s wife] create such a tight knit post production team that you fall into an illusion that you’re making the film for yourselves. It’s almost a sad day when it’s released and belongs to everyone else.”

Originally written for Digital Video magazine / CreativePlanetNetwork.

_________________________________

Needless to say, Gone Girl has received quite a lot of press. Here are just a few additional discussions of the workflow:

Adobe panel discussion with the post team

PostPerspective

FxGuide

HDVideoPro

IndieWire

IndieWire blog

ICG Magazine

RedUser

Tony Zhou’s Vimeo take on Fincher 

©2014 Oliver Peters

The FCP X – RED – Resolve Dance

df_fcpx-red-resolve_5

I recently worked on a short 10 minute teaser video for a potential longer film project. It was shot with a RED One camera, so it was a great test for the RED workflow and roundtrips using Apple Final Cut Pro 10.1.2/10.1.3 and DaVinci Resolve 11.

Starting the edit

As with any production, the first step is to properly back up and verify the data from the camera and sound cards. These files should go to redundant drives that are parked on the shelf for safe keeping. After this has been done, now you can copy the media to the editorial drives. In this case, I was using a LaCie RAID-5 array. Each day’s media was placed in a folder and divided into subfolders for RED, audio and other cameras, like a few 5D shots.

df_fcpx-red-resolve_4Since I was using FCP X and its RED and proxy workflows, I opted not to use REDCINE-X Pro as part of this process. In fact, the Mac Pro also didn’t have any RED Rocket accelerator card installed either, as I’ve seen conflicts with FCP X and RED transcodes when the RED Rocket card was installed. After the files were copied to the editorial drives, they were imported into an FCP X event, with media left in its original location. In the import setting, the option to transcode proxy media was enabled, which continues in the background while you start to work with the RED files directly. The camera files are 4K 16×9 .r3d files, so FCP X transcodes these to half-sized ProRes Proxy media.

df_fcpx-red-resolve_1Audio was recorded as double-system sound using a Sound Devices recorder. The audio files were 2-channel broadcast WAV files using slates for syncing. There was no in-camera audio and no common timecode. I was working with a couple of assistant editors, so I had them sync each clip manually. Instead of using FCP X’s synchronized clips, I had them alter each master clip using the “open in timeline” command. This lets you edit the audio directly to the video as a connected clip within the master clip. Once done, your master clip contains synced audio and video.  It functions just like a master clip with in-camera audio – almost (more on that later).df_fcpx-red-resolve_9

All synced clips were relabeled with a camera, scene and take designation, as well as adding this info to the camera, scene and take columns. Lastly, script notes were added to the notes column based on the script supervisor’s reports.

Transcodes

df_fcpx-red-resolve_6Since the post schedule wasn’t super-tight, I was able to let the transcodes finish overnight, as needed. Once this is done, you can switch FCP X to working with proxies and all the media will be there. The toggle between proxy and/or optimized-original media is seamless and FCP X takes care of properly changing all sizing information. For example, the project is 4K media in a 1080p timeline. FCP X’s spatial conform downscales the 4K media, but then when you toggle to proxy, it has to make the corresponding adjustments to media that is now half-sized. Likewise any blow-ups or reframing that you do also have to match in both modes.

df_fcpx-red-resolve_2The built-in proxy/optimized-original workflow provides you with offline/online editing phases right within the same system. Proxies for fast and efficient editing. Original or high-resolution transcodes for finishing. To keep the process fast and initially true to color decisions made on set, no adjustments were made to the RED files. FCP X does let you alter the camera raw color metadata from inside the application, but there’s no real reason to do this for offline editing files. That can be deferred until it’s time to do color correction. So during the edit, you see what the DoP shot as you view the RED files or the transcoded proxies.

df_fcpx-red-resolve_3We did hit one bad camera load. This might have been due to either a bad RED drive or possibly excessive humidity at that location. No matter what the reason, the result was a set of corrupt RED clips. We didn’t initially realize this in FCP X, and so, hit clips that caused frequent crashes. Once I narrowed it down to the load from that one location, I decided to delete these clips. For that group of shots, I used REDCINE-X Pro to transcode the files. I adjusted the color for a flatter, neutral profile (for later color correction) and transcoded full-resolution debayered 1080p ProRes 4444 files. We considered these as the new camera masters for those clips. Even there, REDCINE-X Pro crashed on a few of the clips, but I still had enough to make a scene out of it.

Editing

The first editing step is culling down the footage in FCP X. I do a first pass rejecting all bogus shots, like short clips of the floor, a bad slate, etc. Set the event browser to “hide rejected”. Next I review the footage based on script notes, looking at the “circle takes” first, plus picking a few alternates if I have a different opinion. I will mark these as Favorites. As I do this, I’ll select the whole take and not just a portion, since I want to see the whole take.

Once I start editing, I switch the event browser to “show favorites”. In the list view, I’ll sort the event by the scene column, which now gives me a quick roadmap of all possible good clips in the order of the script. During editing, I cut mainly using the primary storyline to build up the piece. This includes all overlapping audio, composites, titles and so on. Cutting proceeds until the picture is locked. Once I’m ready to move on to color correction, I export a project XML in the FCPXML format.

Resolve

df_fcpx-red-resolve_7I used the first release version (not beta) of DaVinci Resolve 11 Lite to do this grade. My intention was to roundtrip it back to FCP X and not to use Resolve as a finishing tool, since I had a number of keys and composites that were easier done in FCP X than Resolve. Furthermore, when I brought the project into Resolve, the picture was right, but all of the audio was bogus – wrong takes, wrong syncing, etc. I traced this down to my initial “open in timeline” syncing, which I’ll explaining in a bit. Anyway, my focus in Resolve was only grading and so audio wasn’t important for what I was doing. I simply disabled it.

Importing the FCPXML file into a fresh Resolve 11 project couldn’t have been easier. It instantly linked the RED, 5D and transcoded ProRes 4444 files and established an accurate timeline for my picture cut. All resizing was accurately translated. This means that in my FCP X timeline, when I blew up a shot to 120% (which is a blow-up of the 1080p image that was downscaled from the 4K source), Resolve knew to take the corresponding crop from the full 4K image to equal this framing of the shot without losing resolution.

The one video gotcha I hit was with the FCP X timeline layout. FCP X is one of the only NLEs that lets you place video BELOW what any other software would consider to be the V1 track – that’s the primary storyline. Some of my green screen composite shots were of a simulated newscast inserted on a TV set hanging on a wall in the primary scene. I decided to place the 5 or 6 layers that made up this composite underneath the primary storyline. All fine inside FCP X, however, in Resolve, it has to interpret the lowest video element as V1, thus shifting everything else up accordingly. As a result the, bulk of the video was on V6 or V7 and audio was equally shifted in the other direction. This results in a lot of vertical timeline scrolling, since Resolve’s smallest track height is still larger than most.

df_fcpx-red-resolve_8Resolve, of course, is a killer grading tool that handles RED media well. My grading approach is to balance out the RED shots in the first node. Resolve lets you adjust the camera raw metadata settings for each individual clip, if you need to. Then in node 2, I’ll do most of my primary grading. After that, I’ll add nodes for selective color adjustments, masks, vignettes and so on. Resolve’s playback settings can be adjusted to throttle back the debayer resolution on playback for closer-to-real-time performance with RED media. This is especially important, when you aren’t running the fastest drives, fastest GPU cards nor using a RED Rocket card.

To output the result, I switched over to Resolve’s Deliver tab and selected the FCP X easy set-up. Select handle length, browse for a target folder and run. Resolve is a very fast renderer, even with GPU-based RED debayering, so output wasn’t long for the 130 clips that made up this short. The resulting media was 1080p ProResHQ with an additional 3 seconds per clip on either side of the timeline cut – all with baked in color correction. The target folder also contains a new FCPXML that corresponds to the Resolve timeline with proper links to the new media files.

Roundtrip back into FCP X

Back in FCP X, I make sure I’ve turned off the import preference to transcode proxy media and that my toggle is set back to original/optimized media. Find the new FCPXML file from Resolve and import it. This will create a new event containing a new FCP X project (edited sequence), but with media linked to the Resolve render files. Audio is still an issue, for now.

There is one interesting picture glitch, which I believe is a bug in the FCPXML metadata. In the offline edit, using RED or proxy media, spatial conform is enabled and set to “fit”. That scales the 4K file to a 1080p timeline. In the sequence back from Resolve, I noticed the timeline still had yellow render bars. When I switched the spatial conform setting on a clip to “none”, the render bar over it went away, but the clip blew up much larger, as if it was trying to show a native 4K image at 1:1. Except, that this was now 1080 media and NOT 4K. Apparently this resizing metadata is incorrectly held in the FCPXML file and there doesn’t appear to be any way to correct this. The workaround is to simply let it render, which didn’t seem to hurt the image quality as far as I could tell.

Audio

Now to an explanation of the audio issue. FCP X master clips are NOT like any other master clips in other NLEs, including FCP 7. X’s master clips are simply containers for audio and video essence and, in that way, are not unlike compound clips. Therefore, you can edit, add and/or alter – even destructively – any material inside a master clip when you use the “open in timeline” function. You have to be careful. That appears to be the root of the XML translation issue and the audio. Of course, it all works fine WITHIN the closed FCP X environment!

Here’s the workaround. Start in FCP X. In the offline edited sequence (locked rough cut) and the sequence from Resolve, detach all audio. Delete audio from the Resolve sequence. Copy and paste the audio from the rough cut to the Resolve sequence. If you’ve done this correctly it will all be properly synced. Next, you have to get around the container issue in order to access the correct WAV files. This is done simply by highlighting the connected audio clip(s) and using the “break apart clip items” command. That’s the same command used to break apart compound clips into their component source clips. Now you’ll have the original WAV file audio and not the master clip from the camera.

df_fcpx-red-resolve_11At this stage I still encountered export issues. If your audio mixing engineer wants an OMF for an older Pro Tools unit, then you have to go through FCP 7 (via an Xto7 translation) to create the OMF file. I’ve done this tons of time before, but for whatever reason on this project, the result was not useable. An alternative approach is to use Resolve to convert the FCPXML into XML, which can then be imported into FCP 7. This worked for an accurate translation, except that the Resolve export altered all stereo and multi-channel audio tracks into a single mono track. Therefore, a Resolve translation was also a fail. At this point in time, I have to say that a proper OMF export from FCP X-edited material is no longer an option or at least unreliable at best.

df_fcpx-red-resolve_10This leaves you with two options. If your mixing engineer uses Apple Logic Pro X, then that appears to correctly import and convert the native FCPXML file. If your mixer uses Pro Tools (a more likely scenario) then newer versions will read AAF files. That’s the approach I took. To create an AAF, you have to export an FCPXML from the project file. Then using the X2Pro Audio Convert application, generate an AAF file with embedded and trimmed audio content. This goes to the mixer who in turn can ingest the file into Pro Tools.

Once the mix has been completed, the exported AIF or WAV file of the mix is imported into FCP X. Strip off all audio from the final version of the FCP X project and connect the clip of the final mix to the beginning of the timeline. Now you are done and ready to export deliverables.

For more on RED and FCP X workflows, check out this series of posts by Sam Mestman at MovieMaker.

Part 1   Part 2   Part 3

©2014 Oliver Peters

More 4K

df_4Kcompare_main

I’ve talked about 4K before (here, here and here), but I’ve recently done some more 4K jobs that have me thinking again. 4K means different things to different people and in terms of dimensions, there’s the issue of cinema 4K (4096 pixels wide) versus the UltraHD/QuadHD/4K 16:9 (whatever you want to call it) version of 4K (3840 pixels wide). That really doesn’t make a lot of difference, because these are close enough to be the same. There’s so much hype around it, though, that you really have to wonder if it’s “the Emperor’s new clothes”. (Click on any of these images for expanded views.)

First of all, 4K used as a marketing term is not a resolution, it’s a frame dimension. As such, 4K is not four times the resolution of HD. That’s a measurement of area and not resolution. True resolution is usually measured in the vertical direction based on the ability to resolve fine detail (regardless of the number of pixels) and, therefore, 4K is only twice the resolution of HD at best. 4K is also not sharpness, which is a human perception affected by many things, such as lens quality, contrast, motion and grading. It’s worth watching Mark Schubin’s excellent webinar on the topic to get a clearer understanding of this. There’s also a very good discussion among top DoPs here about 4K, lighting, high dynamic range and more.

df_4kcompare_1A lot of arguments have been made that 4K cameras using a color-pattern filter method (Bayer-style), single CMOS sensor don’t even deliver the resolution they claim. The reason is that in many designs 50% of the pixels are green versus 25% each for red and blue. Green is used for luminance, which determines detail, so you do not have a 1:1 pixel relationship between green and the stated frame resolution of the sensor. That’s in part why RED developed 5K and 6K sensors and it’s why Sony uses an 8K sensor (F65) to deliver a 4K image.

The perceived image quality is also not all about total pixels. The pixels of the sensor, called photosites, are the light-receiving elements of the sensor. There’s a loose correlation between pixel size and light sensitivity. For any given sensor of a certain physical dimension, you can design it with a lot of small pixels or with fewer, but larger, pixels. This roughly correlates to a sensor that’s of high resolution, but a smaller dynamic range (many small pixels) or one with lower resolution, but a higher dynamic range (large, but fewer pixels). Although the equation isn’t nearly this simplistic, since a lot of color science and “secret sauce” goes into optimizing a sensor’s design, you can certainly see this play out in the marketing battles between the RED and ARRI camps. In the case of the ALEXA, ARRI adds some on-the-sensor filtering, which results in a softer image that gives it a characteristic filmic quality.df_4kcompare_2

Why do you use 4K?

With 4K there are two possible avenues. The first is to shoot 4K for the purpose of reframing and repositioning within HD and 2K timelines. Reframing isn’t a new production idea. When everyone shot on film, some telecine devices, like the Rank Cintel Mark III, sported zoom boards that permitted an optical blow-up of the 35mm negative. You could zoom in for a close-up in transfer that didn’t cost you resolution. Many videographers shoot 1080 for a 720 finish, as this allows a nice margin for reframing in post. The second is to deliver a final 4K product. Obviously, if your intent is the latter, then you can’t count on the techniques of the former in post.

df_4kcompare_3When you shoot 4K for HD post, then workflow is an issue. Do you shoot everything in 4K or just the items you know you’ll want to deal with? How will this cut with HD and 2K content? That’s where it gets dicey, because some NLEs have good 4K workflows and others don’t. But it’s here that I contend you are getting less than meets the eye, so to speak.  I have run into plenty of editors who have dropped a 4K clip into an HD timeline and then blown it up, thinking that they are really cropping into the native 4K frame and maintaining resolution. Depending on the NLE and the settings used, often they are simply blowing up an HD shot. The NLE scaled the 4K to HD first and then expanded the downscaled HD image. It didn’t crop into the actual 4K native resolution. So you have to be careful. And guess what, if the blow up isn’t that extreme, it may not look much different than the crop.

df_4kcompare_4One thing to remember is that a 4K image that is scaled to fit into an HD timeline gains the benefits of oversampling. The result in HD will be very sharp and, in fact, will generally look better perceptually than the exact same image natively shot in an HD size. When you now crop into the native image, you are losing some of that oversampling effect. A 1:1 pixel relationship is the same effective image size as a 200% blow-up. Of course, it’s not the same result. When you compare the oversampled “wide shot” (4K scaled to HD) to the “close-up” (native 4K crop), the close-up will often look softer. You’ll see defects of the image, like chromatic aberration in the lens, missed critical focus and sensor noise. Instead, if you shoot a wide and then an actual close-up, that result will usually look better.

On the other hand, if you blow up the 4K-to-HD or a native HD shot, you’ll typically see a result that looks pretty good. That’s because there’s often a lot more information there than monitors or the eye can detect. In my experience, you can commonly get away with a blow-up in the range of 120% of the original image size and in some cases, as much as 150%.

To scale or not to scale

df_4K_comparison_Instant4KLet me point out that I’m not saying a native 4K shot doesn’t look good. It does, but often the associated workflow hassles aren’t worth it. For example, let’s take a typical 1080p 50” Panasonic plasma that’s often used as a client monitor in edit suites. You or your client may be sitting 7 to 10 feet away from it, which is closer than most people sit in a living room with that size of a screen. If I show a client the native image (4K at 1:1 in an HD timeline) compared with an separate HD image at the same framing, it’s unlikely that they’ll see a difference. Another test is to take two exact images – one native HD and the other 4K. Scale up the HD and crop down the 4K to match. In theory, the 4K should look better and sharper. In fact, sitting back on the client sofa, most won’t see a difference. It’s only when they step to about 5 feet in front of the monitor that a difference is obvious and then only when looking at fine detail within the shot.

df_gh4_instant4k_smNot all scaling is equal. I’ve talked a lot about the comparison of HD scaling, but that really depends on the scaling that you use. For a quick shot, sure, use what your NLE has built in. For more critical operations, then you might want to scale images separately. DaVinci Resolve has excellent built-in scaling and lets you pick from smooth, sharp and bilinear algorithms. If you want a plug-in, then the best I’ve found is the new Red Giant Instant 4K filter. It’s a variation of their Instant HD plug-in and works in After Effects and Premiere Pro. There are a lot of quality tweaks and naturally, the better it does, the longer the render will be. Nevertheless, it offers outstanding results and in one test that I ran, it actually provided a better look within portions of the image than the native 4K shot.

df_4K_comparison-C500_smIn that case, it was a C500 shot of a woman on a park bench with a name badge. I had three identical versions of the shot (not counting the raw files) – the converted 4K ProRes4444 file, a converted 1080 ProRes4444 “proxy” file for editing and the in-camera 1080 Canon XF file. I blew up the two 1080 shots using Instant 4K and cropped the 4K shot so all were of equal framing. When I compared the native 4K shot to the expanded 1080 ProRes4444 shot, the woman’s hair was sharper in the 1080 blow-up, but the letters on the name badge were better on the original. The 1080 Canon XF blow-up was softer in both areas. I think this shows that some of the controls in the plug-in may give you superior results to the original (crisper hair); but, a blow-up suffers when you are using a worse codec, like Canon’s XF (50 Mbps 4:2:2). It’s fine for native HD, but the ProRes4444 codec has twice the chroma resolution and less compression, which makes a difference when scaling an image larger. Remember all of this pertains to viewing the image in HD.

4K deliverables

df_4K_comparison-to-1080_smSo what about working in native 4K for a 4K deliverable? That certainly has validity for high-resolution projects (films, concerts, large corporate presentations), but I’m less of a believer for television and web viewing. I’d rather have “better” pixels and not simply “more” pixels. Most of the content you watch at theaters using digital projection is 2K playback. Sometimes the master for that DCP was HD, 2K or 4K. If you are in a Sony 4K projector-equipped theater, most of the time, it’s simply the projector upscaling the content to 4K as part of the projection. Even though you may see a Sony 4K logo at the head of the trailers, you aren’t watching 4K content – definitely not, if it’s a stereo3D film. Yet, much of this looks pretty good, doesn’t it?

df_AMIRAEverything I talked about, regarding blowing up HD by up to 120% or more, still applies to 4K. Need to blow up a shot a bit in a 4K timeline? Go ahead, it will look fine. I think ARRI has proven this as well, taking films shot with the ALEXA all the way up to Imax. In fact, ARRI just announced that the AMIRA will get in-camera, on-the-fly upscaling of its image with the ability to record 4K (3840 x 2160 at up to 60fps) on the CFast 2.0 cards. They can do this, because the sensor starts with more pixels than HD or 2K. The AMIRA will expose all of the available photosites (about 3.4K sensor pixels) in what they call the “open gate” method. This image is lightly cropped to 3.2K and then scaled by a 1.2 factor, which results in UltraHD 4K recording on the same hardware. Pretty neat trick and judging by ARRI’s image quality, I’ll bet it will look very good. Doubling down on this technique, the ALEXA XT models will also be able to record ProRes media at this 3.2K size. In the case of the ALEXA, the designers have opted to leave the upscaling to post, rather than to do it in-camera.

To conclude, if you are working in 4K today, then by all means continue to do so. It’s a great medium with a lot of creative benefits. If you aren’t working in 4K, then don’t sweat it. You won’t be left behind for awhile and there are plenty of techniques to get you to the same end goal as much of the 4K production that’s going on.

Click these thumbnails for full resolution images.

df_gh4_instant4k_sm

 

 

 

df_4K_comparison-to-1080_sm

 

 

 

 

©2014 Oliver Peters

24p HD Restoration

df_24psdhd_6

There’s a lot of good film content that only lives on 4×3 SD 29.97 interlaced videotape masters. Certainly in many cases you can go back and retransfer the film to give it new life, but for many small filmmakers, the associated costs put that out of reach. In general, I’m referring to projects with $0 budgets. Is there a way to get an acceptable HD product from an old Digibeta master without breaking the bank? A recent project of mine would say, yes.

How we got here

I had a rather storied history with this film. It was originally shot on 35mm negative, framed for 1.85:1, with the intent to end up with a cut negative and release prints for theatrical distribution. It was being posted around 2001 at a facility where I worked and I was involved with some of the post production, although not the original edit. At the time, synced dailies were transferred to Beta-SP with burn-in data on the top and bottom of the frame for offline editing purposes. As was common practice back then, the 24fps film negative was transferred to the interlaced video standard of 29.97fps with added 2:3 pulldown – a process that duplicates additional fields from the film frames, such that 24 film frames evenly add up to 60 video fields in the NTSC world. This is loaded into an Avid, where – depending on the system – the redundant fields are removed, or the list that goes to the negative cutter compensates for the adjustments back to a frame-accurate 24fps film cut.

df_24psdhd_5For the purpose of festival screenings, the project file was loaded into our Avid Symphony and I conformed the film at uncompressed SD resolution from the Beta-SP dailies and handled color correction. I applied a mask to hide the burn-in and ended up with a letter-boxed sequence, which was then output to Digibeta for previews and sales pitches to potential distributors. The negative went off to the negative cutter, but for a variety of reasons, that cut was never fully completed. In the two years before a distribution deal was secured, additional minor video changes were made throughout the film to end up with a revised cut, which no longer matched the negative cut.

Ultimately the distribution deal that was struck was only for international video release and nothing theatrical, which meant that rather than finishing/revising the negative cut, the most cost-effective process was to deliver a clean video master. Except, that all video source material had burn-in and the distributor required a full-height 4×3 master. Therefore, letter-boxing was out. To meet the delivery requirements, the filmmaker would have to go back to the original negative and retransfer it in a 4×3 SD format and master that to Digital Betacam. Since the negative was only partially cut and additional shots were added or changed, I went through a process of supervising the color-corrected transfer of all required 35mm film footage. Then I rebuilt the new edit timeline largely by eye-matching the new, clean footage to the old sequence. Once done and synced with the mix, a Digibeta master was created and off it went for distribution.

What goes around comes around

After a few years in distribution, the filmmaker retrieved his master and rights to the film, with the hope of breathing a little life into it through self-distribution – DVDs, Blu-rays, Internet, etc. With the masters back in-hand, it was now a question of how best to create a new product. One thought was simply to letter-box the film (to be in the director’s desired aspect) and call it a day. Of course, that still wouldn’t be in HD, which is where I stepped back in to create a restored master that would work for HD distribution.

Obviously, if there was any budget to retransfer the film negative to HD and repeat the same conforming operation that I’d done a few years ago – except now in HD – that would have been preferable. Naturally, if you have some budget, that path will give you better results, so shop around. Unfortunately, while desktop tools for editors and color correction have become dirt-cheap in the intervening years, film-to-tape transfer and film scanning services have not – and these retain a high price tag. So if I was to create a new HD master, it had to be from the existing 4×3 NTSC interlaced Digibeta master as the starting point.

In my experience, I know that if you are going to blow-up SD to HD frame sizes, it’s best to start with a progressive and not interlaced source. That’s even more true when working with software, rather than hardware up-convertors, like Teranex. Step one was to reconstruct a correct 23.98p SD master from the 29.97i source. To do this, I captured the Digibeta master as a ProResHQ file.

Avid Media Composer to the rescue

df_24psdhd_2_sm

When you talk about software tools that are commonly available to most producers, then there are a number of applications that can correctly apply a “reverse telecine” process. There are, of course, hardware solutions from Snell and Teranex (Blackmagic Design) that do an excellent job, but I’m focusing on a DIY solution in this post. That involves deconstructing the 2:3 pulldown (also called “3:2 pulldown”) cadence of whole and split-field frames back into only whole frames, without any interlaced tearing (split-field frames). After Effects and Cinema Tools offer this feature, but they really only work well when the entire source clip is of a consistent and unbroken cadence. This film had been completed in NTSC 29.97 TV-land, so frequently at cuts, the cadence would change. In addition, there had been some digital noise reduction applied to the final master after the Avid output to tape, which further altered the cadence at some cuts. Therefore, to reconstruct the proper cadence, changes had to be made at every few cuts and, in some scenes, at every shot change. This meant slicing the master file at every required point and applying a different setting to each clip. The only software that I know of to effectively do this with is Avid Media Composer.

Start in Media Composer by creating a 29.97 NTSC 4×3 project for the original source. Import the film file there. Next, create a second 23.98 NTSC 4×3 project. Open the bin from the 29.97 project into the 23.98 project and edit the 29.97 film clip to a new 23.98 sequence. Media Composer will apply a default motion adapter to the clip (which is the entire film) in order to reconcile the 29.97 interlaced frame rate into a 23.98 progressive timeline.

Now comes the hard part. Open the Motion Effect Editor window and “promote” the effect to gain access to the advanced controls. Set the Type to “Both Fields”, Source to “Film with 2:3 Pulldown” and Output to “Progressive”. Although you can hit “Detect” and let Media Composer try to decide the right cadence, it will likely guess incorrectly on a complex file like this. Instead, under the 2:3 Pulldown tab, toggle through the cadence options until you only see whole frames when you step through the shot frame-by-frame. Move forward to the next shot(s) until you see the cadence change and you see split-field frames again. Split the video track (place an “add edit”) at that cut and step through the cadence choices again to find the right combination. Rinse and repeat for the whole film.

Due to the nature of the process, you might have a cut that itself occurs within a split-field frame. That’s usually because this was a cut in the negative and was transferred as a split-field video frame. In that situation, you will have to remove the entire frame across both audio and video. These tiny 1-frame adjustments throughout the film will slightly shorten the duration, but usually it’s not a big deal. However, the audio edit may or may not be noticeable. If it can’t simply be fixed by a short 2-frame dissolve, then usually it’s possible to shift the audio edit a little into a pause between words, where it will sound fine.

Once the entire film is done, export a new self-contained master file. Depending on codecs and options, this might require a mixdown within Avid, especially if AMA linking was used. That was the case for this project, because I started out in ProResHQ. After export, you’ll have a clean, reconstructed 23.98p 4×3 NTSC-sized (720×486) master file. Now for the blow-up to HD.

DaVinci Resolve

df_24psdhd_1_smThere are many applications and filters that can blow-up SD to HD footage, but often the results end up soft. I’ve found DaVinci Resolve to offer some of the cleanest resizing, along with very fast rendering for the final output. Resolve offers three scaling algorithms, with “Sharper” providing the crispest blow-up. The second issue is that since I wanted to restore the wider aspect, which is inherent in going from 4×3 to 16×9, this meant blowing up more than normal – enough to fit the image width and crop the top and bottom of the frame. Since Resolve has the editing tools to split clips at cuts, you have the option to change the vertical position of a frame using the tilt control. Plus, you can do this creatively on a shot-by-shot basis if you want to. This way you can optimize the shot to best fit into the 16×9 frame, rather than arbitrarily lopping off a preset amount from the top and bottom.

df_24psdhd_3_smYou actually have two options. The first is to blow up the film to a large 4×3 frame out of Resolve and then do the slicing and vertical reframing in yet another application, like FCP 7. That’s what I did originally with this project, because back then, the available version of Resolve did not offer what I felt were solid editing tools. Today, I would use the second option, which would be to do all of the reframing strictly within Resolve 11.

As always, there are some uncontrollable issues in this process. The original transfer of the film to Digibeta was done on a Rank Cintel Mark III, which is a telecine unit that used a CRT (literally an oscilloscope tube) as a light source. The images from these tubes get softer as they age and, therefore, they require periodic scheduled replacement. During the course of the transfer of the film, the lab replaced the tube, which resulted in a noticeable difference in crispness between shots done before and after the replacement. In the SD world, this didn’t appear to be a huge deal. Once I started blowing up that footage, however, it really made a difference. The crisper footage (after the tube replacement) held up to more of a blow-up than the earlier footage. In the end, I opted to only take the film to 720p (1280×720) rather than a full 1080p (1920×1080), just because I didn’t feel that the majority of the film held up well enough at 1080. Not just for the softness, but also in the level of film grain. Not ideal, but the best that can be expected under the circumstances. At 720p, it’s still quite good on Blu-ray, standard DVD or for HD over the web.

df_24psdhd_4_smTo finish the process, I dust-busted the film to fix places with obvious negative dirt (white specs in the frame) caused by the initial handling of the film negative. I used FCP X and CoreMelt’s SliceX to hide and cover negative dirt, but other options to do this include built in functions within Avid Media Composer. While 35mm film still holds a certain intangible visual charm – even in such a “manipulated” state – the process certainly makes you appreciate modern digital cameras like the ARRI ALEXA!

As an aside, I’ve done two other complete films this way, but in those cases, I was fortunate to work from 1080i masters, so no blow-up was required. One was a film transferred in its entirety from a low-contrast print, broken into reels. The second was assembled digitally and output to intermediate HDCAM-SR 23.98 masters for each reel. These were then assembled to a 1080i composite master. Aside from being in HD to start with, cadence changes only occurred at the edits between reels. This meant that it only required 5 or 6 cadence corrections to fix the entire film.

©2014 Oliver Peters