Process Stories
Posted on November 6th, 2009 by Anthony van WinkleThe two-year development that has created Night Zero was a process of experimentation, not only in the world of cameras and actors but equally so in the world of digital imagery and photo manipulation. The production side gets more credit all around, because its growth and refinement is more immediately recognizable– just compare the lighting, framing, or backgrounds of the pilot or episode one with what we’ve done in episode three or the Sisters vignette. Throughout the process, post-production has always pushed along in the background, diligently taking its work in and turning the graphic novel out without much fanfare.
In truth, post-production is the most variable part of the entire photocomic process, with a day’s work ranging from an hour or two of rendering and page layouts to a ten- or twelve-hour marathon session of layering, color adjustments, masking, and compositing, all to get the looks just right. Many of the advancements we’ve made in the on-set production over the last two years have been partially driven by what happens in post-production, as we learn what (seemingly) minor changes on-camera can make huge differences in the time required and quality returned in post.
Back in episode one, for our first shoot, we were outside with only a one-person diffuser disc to counter the sun’s unrelenting brightness. If a shot contained Marion on one side of the frame and Claire on the other, we would first shoot with Katrina under the diffuser, then with Tamara under the diffuser, then clear the set and shoot the background with an adjusted white-balance level. These three shots were then separately rendered in HDR and composited together, stitching the properly-lit parts of each into one final photo. I’m grateful to say that such is not how we operate anymore… because of the strain that put on post-production, we learned to take control of our lighting with a three-, four-, or even six-point light system, and diffusers enough to capture the entire scene in one take.
The tonemapping process we use to create the unique Night Zero look is a very immature technology, and has more than its share of quirks and weaknesses that we’ve learned during our time. When we setup shots for frame-area and crop-area, we know that having a reflection of a light source in the captured area will ruin the white-point settings of the HDR image, so we can plan for that and keep better consistency between our tonemapped images (which means less re-tweaking and filtering in post). When we deal with action shots or other hard-to-hold setups, we know that the highest of the three exposures is (under standard lighting conditions) the least significant, and should be the first one dropped if undesirable motion occurs between exposures.
No matter what we did, however, there was still an inevitable reality that post-production is time consuming. Merging exposures into an HDR image isn’t quick, and the mandatory Photoshop prompts ensure that somebody has to sit in front of the computer the entire time, clicking here and there throughout the process. To further aggravate matters, an undocumented feature difference between the “regular” Photoshop and the more expensive “Extended” edition is the (in)ability to convert layers to Smart Objects* when in a 32-bit environment, which means that Night Zero lost that time-saving ability when we upgraded to CS4 last year.
* The future of digital photography and photo manipulation is the advancement of so-called “non-destructive” photo editing, and with each new version of Photoshop comes a new set of tools that can be used in the classic (destructive) or newly non-destructive manner. For example, in both digital and physical photography, a classic method of coloring a photo would be to paint very lightly over the image with the color desired, while a non-destructive method would be to lay a colored gel overtop of the photo. Both methods yield the same result, but in one method the original photo is destroyed in the process and in the other method, the original photo remains untouched. In Photoshop these are called “Smart Objects”, and they are original images that can be scaled, colored, or even tonemapped over and over again without ever changing the source material.
This non-destructive ability is essential to the Night Zero workflow. Going from three source exposures to the comic-style image is a two-step process: first, the three exposures are combined into a 32-bit HDR image, and then the 32-bit image is tonemapped. Tonemapping is an effect with dozens of variables, all manner of adjustments and settings that affect the final look of the image. As mentioned above, even little changes in how a photo is shot or where the light hits the frame can make two nearly-identical HDR images with identical tonemap settings look drastically different. By using Smart Objects while we tonemap, when a rendered image looks different from its neighbors we have the ability to go back to the tonemap settings and tweak them to our heart’s content. Every time the settings are changed, it’s a wait while the computationally-intensive effect is applied, but without Smart Objects we would have to go back to the source images and re-render the HDR images every time as well… an additional three minutes every time we wanted to change tonemap settings (typically done 10-15 times per image).
Photoshop comes with a feature called Batch processing, where one can ”record” a series of actions and then automate a workflow that duplicates this series of actions on any number of files. Unfortunately this is a very limited feature for both of Night Zero’s most time-consuming tasks: rendering three photos to HDR cannot be batched because batch only handles one source file at a time, and tonemapping cannot be batched (in standard Photoshop) because 32-bit files (HDR images) cannot be made into Smart Objects. So for the last year, we’ve been manually rendering every single HDR image we’ve taken, and using an obtuse workaround to trick Photoshop into saving them as Smart Objects. The end result was exactly what we needed, but the process was time consuming and mentally draining.
I’ve recently started getting back into programming, and some minor adventures back to javascript brought to my attention the powerful big-brother of the Batch function: Scripts. By utilizing advanced scripting and unlimited functions, programs can be written to control Photoshop way beyond what can be “recorded” into a Batch job. I was very rusty and had never written for Adobe software before, so it was slow going at first, but sure and steady I began to mold Photoshop to my will. It started with a very simple concept… Photoshop knows the names of the files that are used to make the HDR image, so I wanted to get the program to tell me the first file it used. Once it could do that, I got it to use a part of that filename and automatically save the rendered HDR image in the Night Zero naming convention. When that was up and running, I knew I was in for the long haul.
The Photoshop feature that merges photos into an HDR image is itself just a script, from which code can be copied or which can be directly edited. In due time I had different spawns of that script, each customized for different Night Zero uses. My ultimate goal had been accomplished with the master script: submit one file to the script, and it would take that file and the two following it (which, in Night Zero terms, would always comprise the three exposures) and merge them into an HDR file without any questions or prompts. Once rendered, the file was saved with i’s Night Zero name (taken from the first file of the group), and then the 32-bit file was re-opened and saved as a Smart Object using the workaround I’d discovered back when we first upgraded to CS4. That done, the file was to be closed out and the next one opened, the process repeating as much as necessary.
With that one script, the post-production landscape was turned upside-down. Preparing an entire shoot’s worth of photos for tonemapping and layout, a task which previously took 10-15 minutes per photo, could now be initialized in less than five minutes and left to run on its own. True, in an automated environment the time per photo dropped to closer to five minutes than ten, but that’s irrelevant when nobody has to sit and babysit the machine. Mental energies can be spared that labor, and focus on the real challenge of color-matching the different shots and working for a consistent, attractive look across the entire shoot.
Had I developed these scripts when Night Zero first started two years ago, I don’t know how many weeks of my life would have been spared the grueling task of manually rendering photos into HDR, but if you figure 12 minutes per photo, six photos per page, three pages per week…
Most of all, I’m happy that post-production finally received a powerful boost to productivity and an increased capacity for growth it so desperately needed. So many energies had been focused on pre-production and live production that the work of post had been largely taken for granted, despite the draining and tedious work it required. There is new life in post-production, and much more free time to explore it. I see only good things yet to come.
