On Monday, April 11th, RED Digital Cinema debuted the Epic short film, “Tattoo” in 4K at their NAB booth in Las Vegas. When it was introduced, viewers were told “10 days ago, what you’re about to see didn’t exist.” This couldn’t be more true…
Over the past 4 months, I have noticed that discussions regarding a change in traditional post production DI pipelines have been on the rise. This issue is a common topic amongst filmmakers and post production circles, as well as a frequent topic with online discussion forums. The other issue that seems to have come across my desk a half-dozen times in the last 4 months is the notion that “Epic does not work and there is no workflow.” Rather than point out who is making these statements or discuss who is right and who is wrong, I felt that the best contribution I could make is to simply discuss exactly how the Epic “Tattoo” film was made from my perspective. This posting is meant to produce an accurate timeline of events in order for people to better gauge turn-around times, as well as offer technical information in order to help people make informed decisions about production and post production workflow. And since Tattoo happened so fast, I wanted to document some of the parts I still think about before I begin to forget them. Since there were a significant amount of credible people who worked on this film, you can probably contact some of them and verify that the following statements are, in fact, true and not exaggerated.m
Jim tells the ironic roller coaster story of the conception of “Tattoo” best, so in case he chooses to do so, I’ll leave that part out. What I will say is that I received a phone call at 3PM on Friday, April 1st from producer David Blocker. This was the moment in which I was informed this project was going to happen. David and I went over the challenge brought forth:
April 1st: green light the production
April 2nd: assemble the team
April 3rd: write the film
April 4th: rent the gear & rehearse the actors
April 5th: dress the sets
April 6th: shoot day 1
April 7th: shoot day 2
April 8th: edit the film
April 9th: color correct and mix the film
April 10th: deliver to Las Vegas
I knew this ridiculous idea was somehow going to get done one way or another (that’s just how things work at RED), I just didn’t know how. So I signed Li up and headed to RED Studios for a meeting. When I arrived, I quickly learned exactly how this project was going to be a success: Bill Paxton. Passion doesn’t even begin to describe this guy. Bill is a leader in every sense of the word. I asked Jim what the story was going to be and he said “You have to ask Bill - he tells it way too well.” Bill immediately yanked his phone away off his ear and began pitching me the story and acting it out as it unfolded (all the while someone is apparently on hold, I guess). I began to see pictures in my head and knew that this would not only be an epic production, but that Bill was going to be a total blast to work with. After pitching the ideas still freshly shaping in his mind, Bill cut to the chase, starts a hurried walk towards Jim’s office, and begins pitching people he wants me to contact for editorial, sound and music, etc.
After I left the meeting, I ran into Nate Heartt outside and saw him loading some gear. I said to him “Nate, I don’t know if you heard, but I’m pretty sure there is a tornado forecast to touchdown right where we are standing in about 3 days…” Nate looked up to the clouds and shook his head in agreement. I remember thinking that would be the last good night of sleep I would get for a week. I was more right than I knew.
Over the course of the next 3 days, the sets started rolling in, and the crew list grew to a near perfect team of professionals and volunteers that really made Tattoo come to life.
DIGITAL ACQUISITION SUPERVISION
Part of my job is sometimes referred to as a “Digital Acquisition Supervisor.” I think Dino actually came up with the name, but the concept is simple: when you take filmmaking talent with decades of experience and embark on a production that shoots on a computer instead of a camera, you need to have someone take responsibility for the overlapping processes of capture and post. This incorporates both creative and technical issues, each of which change (in my experience) as often as week-to-week.
For example: From a creative standpoint, it was decided to use two different optical approaches to the film. The film has 2 locations - one present day and one past. The present day footage was shot clean. The light levels were scarily dark and moody and the color tones lived in deep shadows of green and blue. The past footage was shot with filtration - Tiffen low contrast and black promist were used to accent the use of practical lighting, warm tones, wood, and metal. These two completely different looks worked perfectly at the hands of DP David Devlin. It looked so good on set, some might think this would make for easy color correction…but David had a lot of ideas for DI yet to be done. From a technical standpoint, while evaluating David and Bill’s schedule, it was pretty clear that data would need to be managed and prepared for editorial in near real-time. So we decided to setup editor Rob Brakey with an offline by at Light Iron, which is just 4 blocks north of RED Studios in Hollywood. Remember, we only had 2 days to shoot, 1 day to edit, and 1 day to do a 5.1 mix and 4K DI at the same time.
This is where the context of workflow come into play. Unlike tape based digital capture, file-based capture enables users to achieve near-instant backup, evaluation, color correction, processing, and deliverables without a significant brick and mortar infrastructure. Many people know about our OUTPOST systems, which are on-set processing labs. Tattoo would make for a perfect use of an OUTPOST on set, but since Li is only 4 blocks away, we decided to process on an OUTPOST cart at Li. Below is a breakdown of the process:
1. Shoot Epic 5K 2:1 @ 23.98 | 5120x2560 with A and B cameras
2.Bill Paxton directs for digital his first time
3. David Devlin mixes up different filtration for the past vs. present locations
4. Dan Duran runs QTake via SDI 1080p out from camera
5. Monitor QTake playback on 24” JVC
6. Camera shoots to 128GB SSD Mags
7. Location sound mixer, Shawn Holden, captures BWF files to compact flash
8. Justin Jones makes the initial checksum and copy on set
9. Runner takes SSD and CF cards directly to Li for 2nd checksum, copy and processing
10. Colorist Ian Vertovec sets initial look in REDCine-X (443) with Rocket
11. OUTPOST operator Aaron Kroger adds 2.40 matte and syncs audio in REDCine-X using audio auto-match
12. Files are rendered out as ProRes LT (clean, no window) for editing in Final Cut Pro directly to Li internal network
13. Editor Rob Brakey imports clips as they are processed out of REDCine-X for offline
After a full card (20 minutes of footage) is taken out of the camera, the timeline of processing was approximately as follows:
Download onset = 8 minutes
Checksum on set = 8 minutes
Shuttle to Li = 5 minutes
Download & checksum at Li = 8 minutes
Set color, framing, and sync = 5 minutes
Render to ProRes and deliver to FCP = 18 minutes
TOTAL: 52 minutes
That means that on average, from the moment the SSD was full, it was under an hour before the editor was actually working with footage including time to sync, frame and color. This also means that within an hour of capture, there were 3 copies in two locations: 1 spinning disk drive with checksum, 1 raid protected drive with checksum and the SSDs. Because the first day of shooting was over 16 hours long, we did this from about 9AM until 3AM without stopping. The important thing to take away from this “shot clock” is not only is this a best-practice scenario, but turn-around times like this should be the norm for file-based capture on Epic.
• RED is capable of being downloaded approximately 2x faster than real time to backup drives on site
• RED is capable of being rendered to every flavor of AVID or Quicktime format in real time, or even slightly faster than real time
• REDCine-X is a free application that can sync audio, color correct with the R3D Raw files, playback images and export just about every modern codec in use today
• REDCine-X is available online and runs on both Windows and Mac OSX. This means RED has put the equivalent of an SR deck for playback and processing in the hands of everyone with a computer.
After wrapping on Day 2 (this was now Friday morning), Bill came to Li after 3 hours of sleep. He looked surprisingly ready to work and says to me”when you’re going to shoot and edit in the same day, always bring your pillow” -which he did. Honestly, we only caught Bill stealing a short nap or two throughout the day. He had been up for nearly 3 days and his energy and direction was as clear as the first day we met. By the time he came in for cutting, Rob had the entire film assembled and was ready for a directors cut. This was also the first time Bill had really had a chance to look at the footage. The bay we had for Rob has a 50” Panasonic series 11 plasma and ProRes always looks amazing on a plasma. By Friday at 7pm, Bill had locked picture-amazingly we were somehow ahead of schedule!
Cinematographer David Devlin was eager to start setting looks - that night. So we started the DI instantly after Rob and Bill finished picture lock. Rob mentioned to me laster that “this was the fastest and easiest turn-over turn-over ever.” Here is how the turn-over looked:
1. Rob finishes cut and collapses timeline down to 2 tracks (A and B roll elements)
2. Rob exports OMF files for sound mix (remember all audio was synced with hero iso 24bit BWF files during original transcode)
3. Rob exports an H.264 reference movie for sound and picture
4. Rob exports an XML of the timeline out of Final Cut Pro
5. XML is converted to AAF via Automatic Duck - this trick allows us to avoid EDLs which are far more limiting when it comes to opticals, effects, multiple layers, speed offsets, etc,)
6. AAF is loaded into Quantel Pablo and original R3Ds are mounted as well
7. Pablo imports Epic R3Ds with Rocket in 4K - with Rocket support, we can bring the elements into Pablo as 4K or 5K at about 12 frames per second
8. Kelvin and Tint are maintained. For this film, all other meta-data fields are “zeroed” out and the curve of REDLogFilm is applied during import
8. Matt Blackshear conforms the list (which is 9 minutes) and checks it against the the H.264 reference (this whole process takes about 30 minutes)
9. After 30 minutes of work, the list is in Pablo in 4K, H.264 provides the audio guide track, and we’re ready for grading
That night, Jarred, Deanan, Graeme, David, and Bill worked with colorist Ian Vertovec to start the 4K DI. This initial session went for about 4 hours to start setting looks and building an overall style to the film. At Light Iron, we currently use Christie CP2000’s for our grading. While this DLP can only display 2048x1080, the Pablo is working in 4K and scaling down the output only. This is important because all the work, shapes, and manipulation is actually happening on 4K files, not proxies. When we QC the work and play it back, we are watching the actual 4K files. Some systems work with proxy files when resolutions are higher than 2K. This is one reason we prefer Pablo some other systems.
When we do a film project, we will work in log space. Last year we did 2 films that were shot on 35mm and transferred to log DPX. When doing this, we need to put our projectors in P3 space with a print emulation that matches the print we will be film recording to upon completion. When we first started loading in RED footage in summer of 2007, we initially experimented working with the same approach. At that time, REDLog was the only log curve RED offered, and we used that as a starting point with our film LUT for color correction. After working this way on the first RED reel (RED 100), we realized that the images were not behaving as well as they seemed to in RED Alert. Every time we used RED Alert, we were looking at files in Rec709, full code range. After talking with Graeme, we confirmed our suspicions that working with RED files does not require (or prefer) a film LUT to coincide with a traditional LOG workflow. So we turned off the emulation LUT, left log space, and reserved that workflow for formats that were designed for log grading. This is an important step in taking RED files from good to great. Tattoo, The Social Network, Atlas Shrugged, Haywire, Manure, S. Darko, and many others are good examples of how we find working with RED files natively with the REDLogFilm curve applied in a gamma-corrected setup (often referred to as “linear”) such as Rec709 or DCI P3 will yield better results than working through a log workflow such as log print emulation. Once you apply the REDLogFilm curve to your files and start working in 709 or P3, the most common reaction we get from our clients including David Devlin is “Wow. That looks great!” While there is a large number of people who feel this workflow is not in the best interest of digital intermediate, we find that it’s less dependent on color talent, rather more dependent on your philosophy. While most people can agree film makes for a beautiful (albeit cumbersome) format to photograph with, most people will also agree film makes a lousy format to distribute with. Today’s estimations are that nearly 12,000 digital projectors have been installed in North America (of a possible 40,000). 10 years ago that number only a few hundred. With the rapid growth of digital distribution, including the start of 4K projection, along with its superior image quality, digital exhibition is now where the largest percentage of viewers will see a project. So the philosophy is simple: favor the format that will exhibit to the largest audience. Since Tattoo was shot digitally in 5K and will exhibit exclusively digitally, twisting the color matrix to a log print emulation workflow will only limit its potential. While everyone reserves the right to work in the place they are comfortable with, I believe that a gamma corrected DI workflow is the one that will become a regular standard for grading, monitoring and distribution. This is a stark contrast to the companies that have pioneered a working log workflow for film acquisition and print exhibition, which is why one should use both methods, depending on the project. The notion that all images should fit through the same color pipeline is same notion that a single car should be used to commute to work, race on a track and plow snow. Tattoo is a good example of letting the Epic camera perform to its optimal settings and give the DP and colorist the fastest and simplest way to be creative. In Tattoo, there is an average of 8 layers of unique qualification in any given shot. Images such as the long steadicam shot and the old man’s walk form the stove to the window are examples of shots that have more than 20 layers. In fact, when I asked Ian how many layers were in the initial shot of the saloon, there were so many that he wasn’t able to count! This demonstrates the precision of modern DI. DPs such as David Devlin understand this, and when you mix the talent that was available on Tattoo with the latest tools like Epic and Pablo, the word “compromise” is rarely uttered.
And that’s really what it comes down to. There are so many exciting things happening today that we all essentially share the same opportunities to make projects exactly the way we see them in our dreams. So long as we are open to new tools, talent and technology, the need to make creative compromises will continue to decline. That is what excites me, and I’m confident that if we or someone else discovers a new and better way to make a film, we will explore it in a heartbeat. While I don’t think there are any definitive right answers, I definitely can say we have explored and identified wrong ones. That is also what this post is meant to do. I fully understand that people will criticize this process and this workflow. Go for it! Point out my short comings so I can try to improve them next time! Many may have a better way to do the same thing - perhaps light it differently - perhaps shoot it differently - choose a camera they prefer - a workflow they prefer - or color science they prefer - but Tattoo represents exactly what this team of filmmakers wanted to achieve, which is the point. But I will take one last risk: it is my professional opinion that given the timeline and the limitations that we faced 10 days before we had to hand-deliver a product to Las Vegas for exhibition, Tattoo is a film that was executed with near perfection. The Epic camera, even in the hands of people like Bill and David that have never seen it, delivered exactly what we all hoped for without compromises. Thanks to RED’s simple but effective tools, both in production and post, Tattoo was delivered on time, in 4K and I have yet to hear one excuse for quality from any of the top-notch names that appear in the credits.
There is one last button to this story: After a long all-nighter of color correction on Saturday, April 9th into Sunday, April 10th, Katie Fellion, Ian, Kevin London and I carried with us a G-Tech hard drive of Tattoo that was just under 400 gigabytes. We were able to conform, color correct, output and deliver this film to a projector over 200 miles away in 2 days. But after being awake ourselves for nearly 3 days (even 4K credit scrolls take a while to render) we knew we were unable to safely make the drive to NAB. Showing their support, RED decided to send us a driver to take us to Vegas that Sunday morning for delivery to Hugo Cargnelutti who was eagerly awaiting our drive for loading into the Clipster for playback. At 8AM a bus showed up to drive us, only that bus turned out to be a party bus that still smelled like the night before. Complete with a stripper pole, the Light Iron team attempted to sleep on alcohol-stained seats while strobe-light floors and ceilings flashed all the way to Las Vegas. On any other day, that would have seemed strange.
Close to 4,000 people came to the booth and experienced the first project shot and exhibited in 4K on the RED Epic camera. The files were played as uncompressed 4K files out of a Clipster and handed over to RED for RED RAY encoding at 15mbs. Suggesting to viewers that “10 days ago what you are about to see didn’t exist” was an understatement. Epic does work, it is stable and makes incredible pictures. And Epic’s workflow is the most streamlined and readily available cinema pipeline available. If that were not true, no one would have seen the film. And perhaps the best litmus test is that Bill, David, Rob and David Blocker have never used this technology and have since become complete advocates. So much that their showing Tattoo off to their high-ranking friends as I write this. Tattoo is not the biggest project I’ve ever worked on, however it is absolutely one of the most rewarding. So long as most viewers get lost in the story and don’t worry about debayering, cameras, lenses and color science, I’d say we all did our jobs. After receiving that feeling, I slept better than I had in a long time.