Frozen Subject with Moving Camera: An Inexpensive Way of Doing This Cool Effect as Seen in the Matrix

by Robin Liss
Published on Sep 22, 2002 12:00 AM



Ever seen those great scenes in commercials or in the movie The Matrix where the action, and what seems like time itself, is frozen while the camera still moves? This neat effect has been around for a few years but typically it takes tons of money to produce. The way that this effect is achieved is by putting many still cameras, even hundreds, in an array where each camera is a small step away from the previous one. Often the cameras encircle the action and are bent in the shape of the movement. Professionals use $1000 still 35 mm cameras in their arrays, but I decided to try it out with disposable camera.


The first step is making a frame for the cameras so they can all encircle the action. First, I bent my copper wire into a semicircle. The larger the diameter the better. Using this rough template I cut a long piece and two short pieces of wood. I also got three six-inch pieces and drilled holes in them to allow the pipe to fit through them. I screwed the wood together, then screwed the six-inch pieces on the wood with the copper pipe in them. If you're doing this at home, make sure the copper pipe is perfectly aligned and the way you want it.

Now that the frame was assembled, I made the camera setup. You have to trigger each camera at the same time, and there is no plug to plug in, so our solution has to be mechanical. I took a nail, with the head on the shutter, and rubber-band it and duct taped it to the camera. I put another nail under it so it does not press down on the shutter. There was enough pressure on the nail that when the second nail was removed, the first one hit the shutter. I then used clamps to attach the cameras to the brass bar. I let them hang to align them, then attached them once they were equally spaced. Next I attached nylon thread to each second nail and duct taped the thread on. I run all the threads to the back of the frame using eye hooks to guide them smoothly. I made sure no threads crossed and that all threads were running smooth. Then I ran them through a tube.




I estimated that the camera shutter speed is around 1/250th of a second, or not quite that fast. Because these shots are stills, it's best to have a high shutter speed so there is no blurring. One of the possible problems with the rig is that since each camera has a different position, each camera's blur is going to be different for horizontal motion. It's best for this reason to have vertical motion for these shots whenever possible.


I attempted many dry runs to get the cameras to fire at once, with out winding them. After working it out so all cameras fired I tried my first live run, which I think was pretty successful. Due to the angle of the cameras, I could not fire camera number 15. Out of the 14 that were fired, 9 took a picture. Four of those did not take a picture, apparently, because there wasn't enough pressure on the shutter. Only one, the center camera, did not take a picture due to misalignment. I solved the other non-fires by putting more rubber bands on them.

On the second try the misaligned one worked. The misalignment was because I was up above the rig instead of behind it, but I solved this problem by moving behind the rig. Although four cameras did not fire on the second try, I believe that on the second try it was more of a positional problem. The cameras that didn't fire were both on the outside and on the inside. The inside two I think did not fire because they were too close to the string, so there was not enough movement of the pin. I think there might have been too much friction with the outer two.

I tried solving these problems by moving the pins just to the brink of coming out, and by moving them very close to the shutter so the nail was very close to hitting the shutter before being pulled. I also changed the position at which I set the actual firing nail. I moved it right over the shutter instead of a little ahead. Earlier I was putting it a little ahead because it would move back but I think, now that I am more behind the apparatus, that I'm not running into that problem any more.



On the third day of my attempts, I finally got the machine to work. I got a bowl of minicarrots and kept having my subject drop them. I fired the cameras with real pictures each time. The first time three cameras didn't go. I went through and fixed those cameras by making sure that the nail head was over the shutter and by adding rubber bands. The second time only a single camera didn't shoot. It was the same camera that hadn't gone in the previous trial, and it was giving me a lot of problems. Instead of ruining an entire shot, I decided to just replace that camera and I put in a new one. After a few more tries and more tweaking, I got all 14 cameras to fire at the same time with carrots. Then I got them to fire again, right after the first success, for a cake.

It seems that with this system you just have to try with pictures, adjust and try again until you get all to fire. I think that the most common problems were misaligned firing nails and too little pressure.

Next I scanned the pictures into a computer. The first trial I did was just to put the pictures in the editor one after another, give them frame lengths of 2 and watch them play. It wasn't pretty, I knew before this that I would have to adjust the pictures, but it was good to do this to get a general idea of what had to be done. I tried using the moving path feature in Media Studio to move the pictures, but it was very hard.

Instead of using the moving path feature, I developed this method for aligning them. I opened the middle picture, the frontal one, in Photoshop as a reference point. I made it a layer. Then I pasted the next one on top of it and placed it correctly. I also adjusted the colors, it seemed that the color differences in the pictures was a difference in hue. I pasted each sequential picture in the Photoshop image and placed it correctly until I got to the last one. Then, since I only had the second half of the pictures in Photoshop, I did the same with the first half, using the same reference middle picture. After each picture was placed I saved the image as a bitmap, so I had 14 different bitmaps with the correct image in front.

I then pasted these 14 bitmaps into MediaStudio and they were much better. I chose to keep each picture for 4 frames initially. My first trial was with just one picture after another; this worked, but was a little choppy. Next I made the pictures overlap (so each picture was 6 frames with 2 overlapping), and put crossfades between them. This didn't do anything because a crossfade needs at least three frames to do anything (100% source A, 50% A/50% B, 100% source B). I then extended everything to the left so each picture was 8 frames long with 4 frames crossfading to the picture behind it and 4 to the picture infront of' it. The 4 frames per picture seemed a little choppy because it's only an effective frame rate of 7.5, but the clip length was 1:22 so I could always speed it up. There was a lot of extra space on the top of the picture where the pictures overlapped so I put a green matte there. Later, I resized the video to get rid of the green matte. I also resized the picture horizontally to get rid of some lost areas.

There were some serious brightness/contrast issues that I had to address in this project. I found one really good picture and then I matched all the other pictures to it. It was hard, however, because it's almost impossible to make a dark picture light. I thought I had enough light for the setup, but I was wrong. It takes a lot of light. Now the clip was at a point where I didn't feel I could do anymore work in MediaStudio editor.

So, this is just the first trial! I think next time I would like to use digital still cameras. Doing so would allow me to preview the images. If we could hack the cameras to work with an electronic trigger system, there would be no problems with the firing mechanism. I think that, in order to get a smooth fly-around, you need at least 60 cameras, so this will be pricey. With any luck, within the next year we can get the funding to try it out. It might, however, cost $3000-$6000, and by the time we have it perfected, the movie industry and television industry will have come up with a whole new effect.

Watch the MPEG Video (771 K)
Watch the Quicktime video (3 MB)