The following is a technical blog written by Cyle Gage the week after shooting wrapped on Having Fun Up There.
Having Fun Up There is a feature-length movie about a 30-somethingyear-old musician guy who has a hard time figuring out what to do with his life. To me, it’s about a central problem of all artistic life: do you let creativity run your life, or is creativity simply a component of your life? Do you become an “authentic” starving artist, or resort to being a day-job-working “sellout” who practices their art on the side? The main character, Mark, is faced with both options through the people in his life. Capturing these personal interactions was key to conveying the intention of the film.
Primary production took place in nine consecutive days, from September 28th to October 6th. Most of the shots were captured using my custom-built shoulder rig (explained later) or on top of a tripod.
THE CAMERA AND THE RAW
I shot Having Fun Up There on the Canon 5D MK III in 1080p24 14- bit RAW using the Magic Lantern firmware hack. The ML firmware was very stable: I believe I used a late September 2013 nightly build. The only time it ever crashed was once when one of my lenses hit the inside mirror, but the fix was simply to disconnect the lens and take out the battery.
To shoot RAW using the ML hack, I bought a two-pack of Lexar 32GB 1000x CF cards, as they’re currently the only reliable card at a speed fast enough to capture the RAW data at 90MB/sec. Four more cards were bought by other members of the crew, for a total of six cards to store footage while shooting.
That’s 192 gigabytes of total card space. Seem like a lot? To put it into perspective, shooting in 14-bit RAW at 1080p24 takes about one gigabyte every ten seconds. A single 32GB card held, at maximum, five to six minutes of footage. And with ML, you cannot play back your footage on the camera, and you cannot record audio using the camera. To make the RAW footage work at 1080p24, the camera is dumping RAW image files from the sensor to the CF card as fast as it possibly can, using up all of the available data-writing bandwidth. This leaves no room for recording audio at the same time. You have to think along the same lines as using film: you roll it, and you can’t review it until you’ve done something with it. But I’ll get back to that in the “Workflow” section below.
During pre-production testing (which was limited – I bought the 5D and loaded up the ML firmware only a week before primary production) I ran into a few issues when running out of card space. Namely, whatever footage being captured when the card ran out of space would be unusable. However, this issue didn’t recur during actual production, so it may have been something wrong with how I was importing footage.
The movie was shot with four primary lenses, but a total of seven lenses were used. Five of them are M42-mount lenses, using a very basic metal ring adapter. I maintain and stake my reputation on using simple old lenses, as they almost always still look great, and they can add a very nostalgic filmic “soft” feel to a shot. You can easily find M42 (“screw”/Pentax mount) lenses at thrift stores, antique malls, and yard sales, usually for under $100 a pop. There are tons of them churning through eBay all the time. Most of my lenses were either given to me by my father (who used them with his 35mm camera in the 70s and 80s) or purchased in antique malls or on eBay for 25 to 60 dollars per lens.
Since I primarily use my camera to shoot video, I didn’t care about the fact that the M42 lenses don’t have any kind of autofocus. Besides, even when I’m shooting just stills, I often prefer manual focus. If you’re thinking that manual focus is too annoying and timeconsuming, I’d suggest immediately loading the Magic Lantern firmware onto your camera. It has an insanely useful feature called focus peaking that shows little red pixels around whatever is in focus. Finding focus is quick and simple: look at the ML live view screen, pull focus, watch the little focus peaking pixels go from blue and green to yellow and red, and bam, you’re in focus. The only time it doesn’t work reliably is when you’re shooting in extremely low light, but it still makes a very good effort.
My Russian “Imperial Stout” Helios 58mm f/2.0 was the primary close-up/portrait lens used during the production. It’s a very tight lens, and built strangely: the aperture control is on the front of the lens, and the focus is on the back. The aperture control has a weird “locking” feature where you can lock the aperture to a certain maximum f-stop, or adjust it smoothly from f/2.0 to f/16. The Russian lens was not supposed to be the primary close-up lens, but something terrible happened…
My favorite lens was given to me by my dad after I found it collecting dust in a closet. It’s a simple 50mm f/1.7 lens, and it’s very smooth with a classic soft and narrow depth of field. I used it a few times during Sexually Frank to create some of the close/shallow compositions. What was most magical about using it was having a full frame sensor that made the 50mm seem just like human vision. However, during the fourth day of Having Fun Up There production, something happened to the lens which caused it to collide with the mirror every time I mounted it to the camera. (I have since figured out that focusing to infinity collides the glass with the 5D MK III mirror.) After using this lens for the first four days, I had to switch it out for the above Russian 58mm lens.
The “Tiny Tim” 35mm f/3.5 was excellent for medium shots that were very well lit. The depth of field on this lens is okay, but not anything stand-out. It proved a solid choice for two-shots or overthe-shoulder framing, where a little bit of a wide angle was needed. I’m going to try to buy a faster 35mm lens, because it’s a pretty solid all-around focal length for a full-frame camera.
My “fatty” 28mm f/2.8 gave me a scare the first time I put it on the 5D MK III because I had not expected it to be so wide. I had used the 28mm a couple of times on a 7D and I had forgotten that the 5D has such a huge sensor. The 28mm was wide enough to be used for big master shots and only had a little bit of spherical distortion along the edges of the frame. At f/2.8, it’s also fast enough to catch some good low-light action for when the lighting setup couldn’t quite meet the task at hand. I also tried to use the spherical distortion to make certain shots a little weird and open, keeping objects tighter to the foreground while leaving the background heavily out of focus.
We used the cheapest lens Canon makes, the 40mm f/2.8 “pancake” lens. I bought it used on eBay in perfect condition for a good chunk under the retail price. The 40mm is a good all-around lens which can be pulled way back for a medium shot or dug in close for a decently shallow close-up. Whenever the camera was stashed away in a bag, it had this lens on, so I could quickly yank out the camera and get good footage when the need struck. As I’ll explain later in the “Workflow” section, the ability to run-and-gun shoot very quickly was key to our production.
I yanked out a 50mm f/1.8 lens for the last couple of days of shooting, just to see how it would compete with the Russian 58mm and the 50mm lenses. It held up pretty well, but it has a bit less of that stylistic narrow depth of field.
Our massive Canon 70-210mm f/4.0 was used only a couple of times for very long, compressed shots, and a few shots overlooking New Bedford, Massachusetts. Overall, the lens is hard to use for indoor work because it requires a lot of room to use. The closest you can get a subject in focus is 9 feet away, and the fastest it can shoot is f/4.0, meaning you need a lot of good light on the subject.
The new, L-series 70-200mm f/2.8 lens was used just once, for one shot, because it was raining and I wanted to use a lens that was water-sealed.
Another lens my dad gave me was the 135mm f/2.8 “sunset” lens, but we never ended up using it because I forgot it at home most of the time. Oh well.
Lighting the movie was done primarily by me with a very basic three-point lighting kit lent to us by Dan Leich (co-cinematographer on Sexually Frank.) Honestly, I don’t even know the names of the lights, but I usually used the practical lighting available at the location. In the kit, he included a double-bulb-socket contraption which allowed a pair of 75-watt clear incandescent bulbs on a C-stand, which was simple and effective. Typically, I’d like to use that and one or two elements from the basic kit. At one location I also used a Kino Flo Diva-Lite kit because it was so dark inside.
Luckily, for one of the major shoots I had the pleasure of having CE Courtney act as gaffer and key grip for me. We managed to wrangle up some better lighting equipment for him, and he did a fantastic job with the available kits. My only hope is that the lighting in those shots don’t look so good that they make all of my shots look terrible by comparison, but it wouldn’t surprise me.
My philosophy with lighting has always been simple: get light on what needs to be seen. As I’ve observed in many of my favorite movies, lighting is deliberate, but it doesn’t need to be obsessive to be good. Lighting is also one of the primary causes of very long, dragging shoots, because cinematographers believe you can’t shoot a scene without at least three lights set up. It simply isn’t true, especially with how great digital cameras look, even in very low light. Scenes shot at ISO 320, 640, or even up to 1600, all look equally great, even when you have to resort to using a lot of practical lighting. Sometimes an actor can just sit in a chair and naturally look good how they’re lit. My standard for lighting is to not mess with what’s already working, even if it may seem too shadowy or too bright. I’m not saying you should adopt the “fix it in post” attitude, I’m simply stating the fact that sometimes academically “bad” lighting can actually look quite realistic and good.
One thing I hate about most people learning cinematography is the insatiable need to spend tons of money on equipment, whether it’s lights or lenses or rigs. I’m a big fan of the shoulder rig, as it can easily capture eye-lines between actors and be positioned wherever a person can go. On Sexually Frank, I used my co-cinematographer Dan Leich’s basic shoulder mounted rig made out of 25 dollars worth of PVC pipe, lovingly known as “The Bazooka.” For Vibes, I built my own replica of this PVC shoulder rig.
For Having Fun Up There, I wanted to build something sturdier and more impressive.
The answer was $45 worth of black steel piping from Home Depot and a $35 camera plate from Calumet Photographic. The full list of materials, so you can build your own:
• 35 dollar quick release video head adapter
• Six half-inch thick, six-inch long black steel pipe from the “plumbing” section of Home Depot. Around two to three dollars each.
• One half-inch thick 10-inch long black steel pipe from the same “plumbing” section of Home Depot, around three to four dollars.
• Three half-inch thick black steel pipe caps. Around one dollar each.
• Three half-inch thick black steel “elbow” pipe fittings. Around two dollars each.
• One half-inch thick black steel pipe coupler. Around one dollar each.
• One half-inch thick black steel “Tee” pipe fitting. Around two dollars each.
• One four-inch by four-inch steel conduit box with half-inch knockouts, with quarter-inch holes. About four dollars.
• One quarter-inch steel bolt, half-inch long. Maybe 10 cents.
• Bicycle handle grip tape, electrical tape, whatever other flair you want. Shouldn’t spend more than ten dollars on these.
• Optional: three-quarter-inch thick steel pipe clamps for bonus stability. one dollar each.
• Optional, but recommended: a one and a half to three pound counterweight to go behind your shoulder. Can be from a basic weight set bought at any sports store.
How do you set it up? Pretty simple. Screw it all together. You need three basic things: two handles for your hands, a mount for the quick release plate to sit on, and a long pipe section for you to rest on your shoulder. I used the bicycle tape for the hand grips and the shoulder padding, and some red duct tape for flair. Here’s how:
Knock out three of the knockouts of the conduit box. How the holes are aligned is very important. The red boxes are the knockouts to punch out. I used a wrench to do the job. The four small holes along the “top” need to be in line from left-to-right, with the second-to-right hole being a quarter-inch in size. That’s where you’ll mount the quick release mechanism.
Next, put the “tee” joint in the conduit box. Take three of the six-inch pipes and join them within the conduit box with the “tee” joint.
For the two pipes coming out of the sides, add elbow joints, then one six-inch pipe for each hand grip, and then end caps.
For the last pipe coming out of the back of the conduit box, put a coupler on it and then attach the 10 inch steel pipe to that.
Add a counterweight to the end of the long pipe and then add the last elbow joint, six-inch pipe, and end cap. You now have the basic rig!
To mount the quick release plate to the conduit box, screw it in with the quarter-inch bolt through one of the small holes on the conduit box, as mentioned above. Tighten it well, and keep a socket wrench with you on set just in case it gets loose. During the nine days of shooting, I only had to re-tighten it once. Make sure you really tighten and twist all of the joints (the elbows and tee joint) with the pipes so that they’re very hard to move on their own. When you’re done, you should be able to mount the camera to the rig via the quick release plate and be able to hold the rig steadily with either handle and not have to worry about it twisting out of control.
I strongly recommend using padded bicycle grip tape for the hand grips, as just clutching onto the steel can get sweaty and annoying. I also strongly recommend using at least padded grip tape for where your shoulder will hold the weight of the rig, as it can get heavy after a few hours of shooting. I’ve never had any bruising from it, but it’s still a lot to hold a shoulder rig for hours at a time. It definitely gave me some back problems, so I’d recommend doing some lower back stretches before you start shooting, especially if you’re shooting in awkward poses for hours at a time.
THE PRODUCTION AND POST-PRODUCTION WORKFLOW
Actually shooting with the Canon 5D MK III in 1080p24 RAW isn’t hard as long as you have the workflow down. There’s no audio recording, so you need to do non-sync sound with some kind of external audio gear. For Having Fun Up There, we simply had the actor (or director) clap in front of the camera as soon as both the camera and audio gear were recording (the hand claps probably hilariously wasted many, many gigabytes of storage.) So you’ll have to sync up the video and audio during editing as your first step.
The biggest hurdle was doing digital transfers on location. At 10 seconds per gigabyte and six cards, we could only hold about 30 minutes of footage at a time, maximum. For some of the longer scenes, this meant one card could only hold one three- to five-minute take. For almost every shoot we had to be offloading footage from the CF cards to an external hard drive whenever we could so that shooting wouldn’t grind to a halt because of a lack of storage. While initially stressful, we eventually got into a good rhythm of digital transfer. If you can dedicate a person to this task, great. Otherwise, just take a minute in between takes to start offloading cards.
Our digital transfer setup was ideal: a new Macbook Air with a USB 3.0 CF card reader and a USB 3.0 1TB external hard drive (the mid- 2012 Macbook Air and onwards have USB 3.0 ports.) Transferring footage from the card to the external hard drive only took a maximum of 7 minutes per card. The basic on location workflow was to record onto a few cards, and begin transferring between takes. Once we ran out of cards to shoot on, at least a few were freed up while it had been transferring during takes. We probably averaged around 15 pages of dialogue and 60 minutes of footage per day, with shooting time between five and ten hours per day. That’s a lot of footage in a short amount of time. Besides the on-location digital transfer situation, shooting Having Fun Up There was a bit more intense than shooting Sexually Frank or Vibes. For SF, we had two 7Ds doing coverage at all times, with a mix of who was the “primary” camera. We didn’t really have the problem of running out of card space, since we were shooting in H264 (Magic Lantern hadn’t even come out yet) onto a large amount of 4 to 8GB cards. SF was also shot over 16 days spread out over several weekends. Vibes was a single-camera shoot, but it was short (script-wise), more controlled, not very complex, and spread out over a few weekends.
Having Fun Up There was shot in nine consecutive days at various primary locations around eastern and southern Massachusetts which featured heavy dialogue scenes, the need for complex lighting, and the desire to increase the compositional and stylistic quality from previous movies. There were often heavy time constraints, usually because of actors’ schedules or the location’s timetable, which meant getting coverage of the scene took priority over more freeform creativity of composition. At worst, we had 45 minutes to shoot three scenes of heavy dialogue in a dimly lit bar, so we used the same three setups for each scene, having the actors switch clothes before we needed to change lighting for each coverage angle. At best, we had a couple hours at a location to shoot one or two scenes.
While the cast and crew of Sexually Frank, Vibes, and Having Fun Up There pride themselves in being able to pull off such ridiculously complex scenes in a very short amount of time, it does come at a cost. The price for run-and-gun shooting with unpaid actors and crew is a limited palette of shot choices and creative freedom. This is not to say that the movie doesn’t look good or feature great framing and visual composition: it definitely does. But I’m left wondering what more could’ve been done with an extra hour here or there at certain locations, or if a group of extras could’ve hung around for a little while longer to be used as necessary background action.
However, I’ll never trade our production style for the lackadaisical approach most amateur films take, where shooting five pages of dialogue takes a day to cover. It’s also important to note that the cinematography is only one piece of the greater puzzle: when we arrived on all of the locations, it was most often the first time we had seen it at all, the first time the actors could meet to rehearse their scenes, and the whole situation was largely unknown until that moment. A lot of mechanics had to be worked out in a relatively short amount of time. Basically, for nine days I had to show up at a never-before-scene location and light it, frame it, and make people look good in it. But at the same time, the actors had to know their lines, rehearse it, block it, feel it, and perform it, all for the first time as well. And the audio had to be calibrated for the space and the intention of the scene. All of this production being done with only limited ability to prepare for it outside of securing the location, which itself wasn’t always reliable. In these terms, the whole movie is itself a miracle for having been shot at all. That’s the reason why we make movies in the first place – it’s a miracle.
Once the footage is shot and put on a storage array for postproduction, the video files at least have to be synced to audio. This can be done manually in Adobe Premiere very easily, with time and patience. The GingerHDR plugin for Premiere seems to be the only way to drag-and-drop the RAW files from the camera into a video editing program. Color grading has the most latitude, as the camera is capturing ridiculously high-quality 14-bit color. This can be done in Premiere or you can import the RAW files directly into DaVinci Resolve as a CinemaDNG file. Note that the RAW files you’re dealing with have virtually no color information included, so they’ll look really muddy when you first preview them. This is intentional. I don’t think the camera even puts color temperature metadata into the RAW files, Magic Lantern is offloading them to the CF card too fast. However, the color information is all there, so color grading is easy.
Once that’s all done, the movie will look pretty good, and hopefully the whole movie will be pretty good.