A few months ago, I sat down to color an odd indie feature from director Adam Pinney. If you pay attention to the indie film world, you may have heard about The Arbalest after its surprise win of the narrative feature jury award at SXSW 16 despite some less than encouraging reviews (variety, indiewire, screendaily). For those of you who weren't lucky enough to catch this film in Austin, the movie is a meticulously crafted period piece about the "inventor" of "the Kalt Cube" (the Rubik's cube of this alternate universe).
When Adam and I started talking about the look of the film, he emphasized how it needed to look period correct (specifically shot on 16mm). He elaborated that he wanted "it to seem like the film is OF that time, not about that time. I want our film to look like it was shot in the period it's representing." Director of Photography Hugh Braselton shot the film on a simple Blackmagic Cinema Camera which, while providing decent amount of dynamic range, presented an overly sharpened image with pale skintones which would need to be corrected throughout the color process.
Adam collected references covering a variety of looks he loved and how they might apply to certain scenes and sequences of the film. References included The Graduate and some color Hitchcock like North by Northwest (despite the fact it was shot a decade before The Arbalest was supposed to take place). In particular the warm skintones that were characteristic of the period is a look we loved. As the film progresses (and time moves on) we move to Godard's A Woman is a Woman and the saturated colors of Weekend and Scorsese's Taxi Driver.
Reference frame from Godard's Weekend
Beyond the tones, Adam made it clear he wanted grain and lots of it. We experimented with both scanned and generated grains from a variety of companies and ended up with a mix of generated grain from Film Convert OFX within Resolve for the 16mm sequences and some Rgrain scans for the 8mm portions.
From a workflow standpoint, I love using Film Convert OFX for generated grain. The grain is very realistic (built of high quality scans), but more importantly allows you to work quickly. It's simple to slap on an OFX node and append it to multiple clips, groups, or even a timeline and customize intensity, size, etc for each scene, angle, or shot. Layering scanned grain is a much more laborious process and, for me at least, the difference just isn't worth it in terms of your final product. It is worth noting that I tend to avoid using the "film color" and "film curves" features of Film Convert as I find the results are often lacking. Furthermore, if you are working in an ACES project, Film Convert should be avoided as the way it handles gamma results in clipped highlights. Regardless, it's well worth the $199 the OFX plugin costs for the flexibility of the grain alone.
Because I'm in NYC and Adam was editing in Atlanta, we had to get a drive with all the media, some sort of project file/XML/EDL, and a reference render shipped to me. HD and now 4K+ projects can have a lot of media. I've worked on five minute music videos that have had well over 1TB of media (I currently have two 30 second spots with nearly 2TB's of accompanying media (somehow) sitting on one of my RAIDs). You can understand how a nearly 90 minute feature can quickly become impractical to ship if you include all the source media. Since the project was edited in Premiere Pro CC on OSX, I had Adam prepare a project managed version of the film to cut down on how much media would be sent. We settled on reencoding to ProRes 4444 with 12 frame handles to give us a little bit of wiggle room while preserving as much of the original data as possible (and keeping file sizes sane).
After receiving the drive, I transferred the media to my working RAID (8 bay 32TB RAID6 connected via Thunderbolt 2), backed up additional media to my server, and stashed the drive for safe keeping. The project opened without any issues and I generated an XML for Resolve which worked nearly flawlessly outside of some reframing issues and some problems relinking to a few still images (which Adam corrected during his final online). This was all checked against the offline reference render provided by Adam from the master project timeline.
It's worth taking a moment here to examine what constitutes a successful period look. There are two things to keep in mind whenever you're imitating a well-known past style: what it actually looks like and what people think it looks like.
Bear with me for a moment here. When we were approaching the grade for Too Many Cooks, director Chris Kelly and I went through dozens of reference shows and intros to find "the look." The short (also shot on Black Magic coincidentally by Adam Pinney as DP) before color was flat and sharp (notice a trend here) and desperately needed to be stylized. While looking through YouTube rips of various intros we were imitating, we were surprised to find that they didn't look like what we "remembered" them to have looked like. Granted, part of that is because these rips were from various sources at varying qualities (as opposed to the "pristine" original broadcast), but the larger issue seemed to be that what we "know" something looks like is shaped by how we remember the time and the similar looks, styles, and images of the period. Our memory creates the look that feels right and when we see what the original actually looks like it can feel "off." We found that we got a more effective "look" by building off of what these intros were supposed to look like than by copying the actual recordings 1:1.
For The Arbalest, we had very similar findings and learned through experimentation that identifying a number of stylistic tropes from the period and imitating them gave us a more convincing period look than simply copying our reference scenes. These tropes were narrowed down to saturated orange skintones, blooming in the highlights, slightly lifted blacks, softness on the high end, boosted highlights, a general warm tone, and heavy grain.
To speed the process up, similar shots were put into groups (which if you aren't using, you need to get on that ASAP) making it fast and simple to grade scenes and later make tweaks as notes came in.
A typical shot's node tree looked like this:
First thing I did was add the log to 709 transform LUT as my last node. Most shots used the Blackmagic provided transform, though I found using the Arri LogC to 709 achieved interesting results for selected scenes and is something worth experimenting with. Note that this technique has a tendency to supersaturate colors so avoid this on anything with bright, intense colors (LED Christmas lights for example).
The first node handled primary corrections which was typically balancing levels (including milking the blacks a touch), correcting hue shifts (lots of green in the raw footage) and warming the footage to the approximate final look.
I then stacked three parallel nodes (for simplicity of layout and visualizing the workflow more than any reason that requires parallel nodes). The first node uses HSL qualifiers utilizing solely the luma qualifier to roughly select highlights with a bit of a soft rolloff to midtones. This node then boosts the gain and a little gamma to increase highlights across the image and make the lighting more dynamic.
The second node in the stack is another HSL Qualifier this time utilizing hue, saturation, and luma to grab skintones (and whatever else is close to that tone in the shot). The skintones are then warmed and saturated further to hit that North by Northwest feeling. Some shots limited this to just skintones with masking while other shots let the qualifier affect everything that fell within the tone throughout the shot (as in the example above).
The last parallel node was again an HSL qualifier with only luma qualifying enabled. This node took the highlights of the image and dramatically increased the "Clean White," "Blur Radius," and "In/Out Ratio" to get a large, soft selection around highlights as seen below:
This selection was then brightened, primarily by increasing lift. Once brightened, the highlights were further smoothed out by decreasing the midtone detail and adding a very slight blur. The effect is a soft bloom around bright areas of the frame that brings to mind older soft lenses prone to halation.
The last step of the process is appending an OFX node to the end of the chain to add grain through Film Convert. I like to do this en masse after the grade is polished and then tweak per group to make sure grain levels are consistent and natural for each shot and sequence.
Beyond this, shots would receive various clean ups (making sure white stayed white and blacks were true), power windows were tracked to brighten up, darken, or fix colors in problem areas, vignettes were added where appropriate, and color tones tweaked to complete the image. I tended to hit these changes after the main look (the nodes mentioned above) was built.
The revisions were the piece were minor and focused on correcting some saturation issues and fixing small conform issues that were caught during the online. The piece was mastered in both DPX and DNxHD 220 10bit at 1080p23.976 and then sent to a facility for conversion to DCP through the standard XYZ transform.
The demands of nailing this stylized look made the piece tons of fun to work on. I was impressed with how far the Blackmagic was able to be pushed and how quickly we were able to push the footage to the "period" look - due in no small part to the excellent production design. The success of the film on the festival circuit has been exciting to watch and I hope it means the film will find larger distribution for everyone to enjoy it.