The build started on March 31st 2014, and my videos run until the main structure was completed, around the 22nd July. I set the camera to record between 5am and 8pm at 1-minute intervals. This meant it produced a minimum of 900 frames each day. It would produce more frames on busy days - around 1200 frames - as it uses motion detection as well to trigger the camera. I could have chosen to switch off motion detection, but I left it on in order to see more detail. So while the timeline might be slightly skewed, in the grand scheme of things it's not making a huge difference by the time I've cut the relevant parts of each film sequence.
If you're only going to watch one of these, watch Part 2; it's the most visually exciting, seeing the structure being built.
I've learnt a few things from doing this.
I used a Bushnell field camera which is normally used for spotting wildlife.
The advantage of this is:
- it's camouflaged (it took forever for the house owners to spot it, tied to a tree)
- batteries last 2-3 weeks (takes 8xAA)
- it's weatherproof
- You can choose between time lapse or motion detection. If you look at my videos, it's doing both; there's a T or M flag in the bottom left that tells you what the trigger was for each frame.
It doesn't, however, handle bright white concrete very well, since it's designed to capture brown animals against green and brown trees. So I found that bright Texas sun on bright concrete and pale wood would wash out badly. I've reduced the exposure of pretty much all the morning-to-mid-afternoon sequences down by a stop and a bit, but the files are all jpg and that's about as good as it gets.
I've also had a few instances where the camera just stops recording, for reasons best known to itself. I _think_ the culprit is a dodgy battery connection; it seems that if you run your fingers along the line of batteries, it's enough to bring it back to life. But it's annoying when you've left it sitting for a few days and have nothing to show for it.
You could use a GoPro and that was my first thought, since they are small and weatherproof. However, the battery life is very short. When you run time lapse on a GoPro, it stays on constantly, it doesn't shut off between frames. So you get the same ~3hour battery life as if you were running video. I have seen folks mod the case by drilling a hole over the power input and then running the camera off whatever power source they have fixed up, but then you're no longer weatherproof. It's also more expensive losing a GoPro than a Bushnell, if someone decides to walk off with it.
If I was doing this again, I'd research my field cameras to see if there's one that is better optimised for bright subjects. The one I have was originally bought to see what kind of critters visit us in the night, so it's just what I had on hand for this job. * see Addendum; below.
I'd also like to get two or three of them to film from different angles, so if one drops out then you still have others. Also, houses are big, and I've lost the right hand end since it doesn't fit into my field of view. There's many days when the builders are doing things around the other side of the house, and I just can't see it. However the tree I have it tied to was about the only place I could put it that was going to remain out of the way and have a good view of the site.
One other thing I've learnt is that you shouldn't be afraid to move the camera. I was married to the fact the camera was in its one spot, and wanted to have the finished video from all the same viewpoint, for continuity. However, we ended up with a big gap of at least six weeks between the outside of the house being finished, and the concrete driveway being poured. I was all set up to get the little bit of pathway that was all the concrete which would be appearing in the shot. Then I spoke to the homeowner about it. He'd been busy; I'd been away and also busy, neither of us had fully thought about what was going on. We decided to move the camera after all, to catch more of the driveway being put down. Of course, if we'd had this conversation earlier, I'd have moved the camera and caught his swimming pool being installed. You live and learn though. As it is; the camera decided not to record the part where they actually laid the driveway! Since a flat white bit of motionless concrete isn't that exciting, I have finished the video with the walls being coated in stone, instead.
As for editing the movie, this takes patience, and hard drive space. 900 frames per day eats up hard drives pretty quickly, so I'm using an external Drobo drive to hold it all. I'm copying each day's worth of images into a separate folder. Lightroom and LRTimelapse are used to bolt everything together.
You have to open each day's directory which will ingest all the frames into LRTimelapse, then click to initialise them, in order to be able to play back that day's sequence. This takes about five minutes, and only then do you know if you've caught anything interesting that day. If nothing happened, repeat with the next day until you find some action. Then you can get into Lightroom and go back and forth with the files until they're corrected the way you need them - you will likely need to alter exposure and/or cropping, as well as deflickering the images. This page on LRTimelapse.com explains the process. Then you can export and render the video sequence. If you have around 1000 images files, this lot will take 20-30 minutes, for each day's video.
Once you have a bunch of daily videos, you can load them into Final Cut or another video editor, and cut them until you have something like the videos shown above.
I wrote most of this blog post around September 2014. I'm publishing it in February 2015. For many reasons, I have been delayed in making part 3, but it's here now. I've also recently become aware of the existence of Brinno time lapse cameras. Their construction camera would have been the perfect tool for this job; it records for months at a time and builds the video in-camera. Oh well! I guess if I do this again, I know which tool I'll use.
Anyone else want to build a house?