How to Create a Hyperlapse From 7 Still Frames
Finding ways to approach shots in new ways is an imperative part of any learning process. Armand Dijcks, a filmmaker out of Rotterdam, The Netherlands recently found a cool way of mimicking hyperlapse through a new technique he calls, Morphlapse.
Armand has shared other techniques with us in the past, namely, the technique of long exposure timelapse photography. If you have yet to check out the post, I highly recommend giving it a read. With this post, what Armand does is walk you through how to approach these types of shots while giving you a few tips to help along the way.
Without further ado, attached is the cool technique presented by Armand.
An experimental hyperlapse shot created out of only seven still images.
The above shot came about while I was in Chicago teaching a workshop. The historic Wrigley building with the Trump tower looming behind it struck me as an interesting subject for a hyperlapse, especially because of the shape of the Wrigley building, which makes it look wider or narrower as you move from one side to the other. Because I didn’t have time to do a proper hyperlapse using a tripod, I decided to do an experiment and do it hand-held. I figured that I could always stabilize it in post.
That proved overly optimistic in hindsight. Sure, I could apply some software stabilization, but at several points during the sequence there were abrupt and very distracting movements. An option would have been to correct position and orientation of all 220 shots manually, but I decided to try something completely different. I took just 7 frames from the sequence, in regular intervals. Since I had 220 images, that meant I selected every 30th image.
After stringing them together into an ultra-short hyperlapse (at 30fps this wasn’t even a quarter of a second of footage), I slowed the resulting clip down to 2% speed in Final Cut Pro, using the “optical flow” setting. Normally used for slowing regular footage down to create slow-motion, optical flow interpolates between the existing frames to create new ones. In this case, since I went from 100% speed to 2%, the algorithm has to make up 50 frames between each pair of existing frames.
Sounds like a bit of a stretch, but the result was a surprisingly smooth hyperlapse. Granted, the shapes of a lot of the buildings were doing a weird kind of morphing that is typical for this method, but the central part of the frame was remarkably stable. In fact, some experimenting showed that it became more stable the less images I used to start with. Of course there was a limit to this, because with too little initial images the morphing effect gets out of control. But I decided that the morphing could also be used as a creative effect instead of trying to avoid it altogether. What’s more is that the whole thing had a kind of slow motion feel to it, rather than the nervous, jittery feel that most hyperlapses have.
So far I’d just been using the unprocessed RAW images though. I figured a more stylized look might work better here, so I did some processing on the images in Aperture to bring out some nice crisp contrasts and details. However, all of that detail proved to make it a lot harder for Final Cut Pro to transition between the individual images, and I was getting some crazy morphing effects again.
This happened mostly on the left and right side of the frame, since that’s where the individual images were most different from each other. The solution was to apply a tilt-shift blur to the initial 7 image sequence, blurring the left and right side of the frame. The blur makes those areas much more uniform. The software has a much easier time interpolating soft, blurry areas than areas with lots of detail. The central part on the other hand stays relatively consistent throughout the 7 frames, so the optical flow algorithm does quite a good job there. Of course the blur has to be applied before the time stretching, otherwise it will have little effect. Applying the blur after the time stretching only hides some of the excessive morphing, but it will still be distracting.
This finally gave me the smooth look I wanted. The blur also helps to create a strong visual focus on the Wrigley building in the center of the frame.
This is a fun technique to create smooth hyperlapse shots with relatively few images, if you don’t mind the morphing effect. There are limitations of course. If large parts of the image change too much over time, there’s no way FCP can create smooth and consistent motion from them. Although I haven’t tried it yet, it might also be worth experimenting with plugins like Twixtor or Kronos, since they’re made specifically for time stretching. As always, it’s a matter of experimenting to see how you can achieve the best results.