It seems like I’ve been spamming my blog about this bullet-time video for months, in reality it’s only been 4 or 5 weeks. I’m sure all my regular readers are dying to see the finished video. If you can’t wait any longer, scroll to the bottom of this post.
Even though it’s been a relatively short period of time, I feel like I have made a journey of epic proportions as far as discovery and trial & error are concerned. Before I started this project I had almost no idea of what would be involved in creating this video.
At first I thought it would be as simple as sticking a few cameras next to each other, pointing them in roughly the right direction and hitting record. Oh how wrong I was. I ended up working out circumferences of circles, comparing the stiffness of plywood and MDF and inventing a booking/tracking system for borrowing cameras. All part of my bullet-time journey.
It felt like I was stumbling around in the dark to begin with. Greg, the college tech guy, and I used 6 cameras placed on a table. The results were rubbish, we didn’t even know if the frames should overlap each other or not. The most annoying thing was, there was absolutely no information about creating bullet-time. Even the mine of information that is the Internet proved pretty useless. There are lots of photos of nice looking camera arrays, but no real useful details like where the cameras should be pointing, how big a radius they should be on and how to make sure they are all framed identically. We had to work all these things out for ourselves over a series of tests that spanned 2 weeks.
Once we’d worked out all the details it was relatively straight forward. The most difficult bit was getting hold of enough cameras. I managed to cobble together 20 in the end. I had 15 550D/600Ds and 5 7Ds. I used 19 18-55mm Canon kit lenses and a 20mm prime. I had to scale all the footage up slightly to match the footage from the 7D with the prime lens.
Alignment was a big problem. I built the rig out of MDF and reinforced it with batons. It was rigid as anything but I still had framing issues. I discovered that the lenses moved when I touched them to adjust the focus. I’d assumed that the lenses were held tight to the camera bodies but they really aren’t. Luckily I got around this in post by lowering the opacity and moving the footage around to line it up. I’ve lost some quality due to having to do that, which is a bit disappointing but there was nothing else I could do.
The edit was a monster task in itself. It took me a week to transfer all the footage from the cards (I had to keep track of which card came from which camera too) onto my computer and compress it. There was 150Gb of raw footage and the final video is 2 minutes long…
Once I’d got the footage into Premiere, the real work began. The first editing day I worked for 12 hours and got 9 seconds of usable footage. It took an age to mark in points on each clip from all the cameras and then work out where I wanted to start the effect from. I had to cut each clip 1 frame longer or shorter than the one below it in order to create the effect. Once I’d got the pan where I wanted it I had to go back to every clip and align it to the previous one so the effect was as smooth as it could be. It was painstaking work but it resulted in some pretty interesting looking timelines.
After getting the raw, full-speed bullet-time clips done, I started running them through After Effects. I wanted to slow them down so you could really see what was going on during the tricks. I was also hoping that it would smooth out the motion a bit, as Twixtor adds frames. I’d planned to do this from the very beginning, so I shot using a really high shutter speed and at 50fps, the max for the cameras I was using. That took a few days and after that, I was finally ready to lay some clips down in Premiere to a music track.
So that, more or less, is my bullet-time journey. Greg and I are planning to make a behind the scenes/how to video. More news on that as it develops but for now, enjoy the video.
Matt Champion: Bullet-Time Unicycling
A massive thank you to everyone who helped make this video possible. Dominic Bennett, Ashley Compton, Emily Cox, Megan Hoang, Monkey Newton, Jake Titmus and Amit Vadgama for kindly loaning me cameras. My Dad for helping out building the rig and most of all, Greg Mullins for being the guy that can get things. Without Greg, this video wouldn’t have happened.
This also sort of marks the unveiling of my new website. I needed one anyway, but I had to produce marketing materials for myself for this unit and I thought it was the perfect excuse. I will still be blogging here, I want to keep my personal and professional sites separate.
Head on over to http://www.eddhawkes.co.uk and check it out. It’s simple and minimalist but I wanted it that way. It doesn’t work very well yet, the links are all fucked up and I can’t get rid of the horrible drop shadow on some of the text. If anyone reading this is good with CSS and is willing to have a look at it, give me a shout!
Until next time, adios!