I think in the end, this had to be faked. I don’t think you could lock the speed of the car so tightly as to keep the animated image from drifting left or right once up to speed.
It would have to be an insanely tight feedback loop, with all sorts of sensors and methods to keep the car at exactly the speed needed!
Posted by Nate Weaver on 11/09 at 07:20 PM
I want to believe.
-MJ
Posted by Matt Jeppsen on 11/09 at 07:27 PM
Updated with making-of video. Believe. 
-MJ
Posted by Matt Jeppsen on 11/09 at 07:46 PM
“I don’t think you could lock the speed of the car so tightly as to keep the animated image from drifting left or right once up to speed.” What if you had a video-tap monitor on the dash, in the driver’s FOV? Then it’s a simple matter of making very minor speed variations to keep the animated pix centered in the frame. No more difficult than holding station on another moving car, really. That’s how I would stage it, at least on a closed course.
If I were really worried about the divided-attention problem, I’d let the driver steer, and have a backseater view the tap monitor and control speed with a remote throttle. But on a deserted straightaway (deserted aside from the “making-of” camera car, which is keeping a healthy distance), I don’t think the workload would be too high for a single driver.
Whether the insurance and bonding companies would agree with me would be a different matter, grin.
Posted by Adam Wilt on 11/12 at 01:31 AM
My name is Peter McAuley. I was the visual effects supervisor on this Subaru spot. Just to set the record straight, it was a 1 Degree shutter. How did we determine that? In an earlier test we shot a brick wall straight on out the side of a van with a 1 meter field of view about 2 meters away with the focal length of about 60 mm on an 24-290 Angenieux Optimo. We travelled at speedometer speeds from 10-70 km/hr in increments of 10 km/hr. We tested 3 cameras. Red MX, Sony F-35, and an Arri 35 LT.
When we looked at results, we knew we had to throw out the Red due to it’s cmos chip and the horrible skew it produced as a result of the rolling shutter and therefore any cmos camera was taken out of the mix. The Arri 35 LT was set to 1 degree and held perfectly sharp on the wall until 50km/hr. The Sony F-35 was quite good as well. It was good to about 25km/hr. No skew issues with it due to it’s CCD sensor. The only issue was that it would only go down to 4.5 degrees. We felt that 20ish wasn’t fast enough to really be exciting. The decision was made to shoot on film with the Arri 35LT. Next on the list in the math and science department was how big should the animation frames be. We wanted to be safe on the speed to ensure sharpness so we settled on 40km/hr.
40k =11.088 m/sec. Our animation is 24 frames a second, so 11.088/24 =.462m. So we had the printer print the vinyl with frames .462m wide. 720 frames X .462 = 362m of vinyl. Built a fence 450m long then attached the vinyl with grommets. Next on the list was how to synch the film shutter on the Arri to the frames of animation so that we wouldn’t see the frame lines in between. Answer: trigger the camera start with a laser.
Searched the internet and found a company in Pennsylvania IFM Effector. Found a laser trip device built in Germany with to output triggers. Had Rudy our engineer wire a few arming switches in a box. Mounted the laser on the truss on the outside of the car. Set the laser to trigger on any thing less than 6m. Then I had the grips mount a 4 X 8 white form core about 5m away from the car. The car would travel down the drag strip and before we got to the animation on the fence the laser would hit the foam core and trip the start of the Arri. Well that solved the start issue but didn’t solve the synching of film shutter to animation frame. I then measured from the edge of the foam core (where the laser tripped) down .462m (the width of one frame) and made 10 increment marking on the ground. Called these P1 P2 etc. The video tap on the Arri sort of emulates the 1 degree shutter, so we could see the frame line on the animation and make position changes on the foam core edge to find the right one that would show us the full animation with no frame lines. Now after we shot the test we discovered ( although Adam Marsden the DP and I predicted it ) that when we compared the video tap recordings with the film dailies that the two images were 180 degrees out of synch from each other. So when the video tap was in synch the film image had the frame line right down the middle. Makes sense really because of the rotating mirror in the Arri. While the image is going to the tap the film is pulling down the next frame of film. Hence the video image is first followed by the film image. So on the day we were actually trying to get the frame line right down the middle on purpose knowing that on film it would be fine. Needless to say production had to put a lot of faith in what Adam and I were doing. The final piece to the puzzle was the constant speed of the car. We bought a military grade GPS that synchs up to 18 satellites at once and has realtime speed readings accurate to .1 of a kilometer. I have been down the road before trying to use the cruise control on vehicles and they are not accurate at all. Andy a third party hotrod car tech plugged his laptop into the WRX and set up a rev limiter on the car. Watching the GPS speed we tweaked the revs until we achieved 40 km/hr. I think we ended up at 2725 rpm in 2nd gear. We figured if we could keep our speed to within .1 k we would be good. All our good takes we achieved speeds of no lower than 39.9 and no higher than 40.1. Now that all the math and science was done the only thing we had to worry about was exposure. All you DPs out there know that at 24 fps and a 1 degree shutter you exposure is -7 and 2/3 stops down from base. We used the Kodak 500T pushed one stop to 1000 with no 80A. On a sunny day that bought us about 4 and a 1/3 on the optimo. God help help us if it was overcast. Of course it was and in fact raining most of the day. So wide open it was. 2.8 This was probably the biggest math and science project I’ve been on in a while. Great fun. Again just for the record the making of left out all the good stuff that techno nerds like us really like.
Peter McAuley
.(JavaScript must be enabled to view this email address)
AXYZ
Toronto
Posted by Peter McAuley on 11/12 at 02:22 PM
Thanks for the detailed behind-the-scenes explanation!
Posted by Adam Wilt on 11/12 at 11:40 PM
Peter, thanks so much for sharing your detailed explanation! Article has been updated to correct the shutter angle info.
-Matt
Posted by Matt Jeppsen on 11/19 at 09:29 AM
Subaru Canada chose us. It was a television commercial for the WRX. Our job was to make it work with their car. I’ll admit it was quite a challenge with the window profile being so small. The size of the animation frame was pretty much predicated by the size of the window. Field of view 101. We actually traced the window and cut a hole in some cardboard and used it in the early tests, to help us get the math right. What a great car. The precision driver we had was really good.
Peter McAuley
Posted by Peter McAuley on 01/06 at 07:43 PM
