This music video was released yesterday:

It's a brilliantly simple idea for a video: a camera continuously travelling through rooms while those rooms are getting smaller and smaller.

Artistically, the metaphor is a powerful one, there's no doubt about it.

Technically, it caught my attention by how seamlessly the transition between the rooms is done. Watching it for the first time, taking the first four rooms as an example, I thought they used Green Screen to superimpose the actor and a bit of CG to retouch some details, but it looked way to perfect (in 3D lingo it means everything was to imperfect looking). By the end of it, I was certain that no green screen nor CG was used - only good old fashioned models, controlled camera motion and take repetition, with a lot of masking and rotoscoping in post-production. According to the director:

“Doing this film with CGI would have been a thousand times easier, but for me, it’s physicality and imperfections are what make it different, and I hope better,” he adds.
"We had to design an entirely new way of moving our miniature camera to get it to fit through the tiny doorways and travel smoothly for such a long distance. We also had 18 rooms all at different scales, so we had to duplicate and shrink every single prop that features in the set, which is obviously was a ton of work."

How was it made?

From what I read and assuming a lot, this is what I think the process was:

  • The first room was build on a 1:1 scale.
  • The adjacent rooms where built on a smaller and smaller scale, keeping the ratio
  • Special attention was given to the camera's movement so everything keeps coherent
  • Multiple takes
  • Lots and lots of masking and rotoscopy

Almost every prop in that first room was rebuilt to scale for each room, which is crazy time consuming!

 Notice the difference in scale in both shots. Notice how almost every prop was reproduced and rebuilt to scale.

Notice the difference in scale in both shots. Notice how almost every prop was reproduced and rebuilt to scale.

Although being time consuming during pre-production, when you break it down, it turns out to be a really simple idea technically, right ?

After getting this all sorted out, the tricky part is to fake the camera movement in a way that you don't notice the difference in scale, unless you use the man as your relative point of reference for scale.

I don't want to pretend I fully understand Einstein's Theory of Relativity, but I will pretend I understand the concept of Space-Time. In a nutshell, Space and time are intrinsically connected, space being represented in 3 dimensions and time as a 4th dimension. In this specific case, it means that in order to keep a linear movement when you down scale space, you need to down scale time in the same ratio.

What is happening - despite it looking like the camera is keeping a constant speed throughout the entire time - is that the camera is actually slowing down from the first until the last room. You want the camera to take the same amount of time travelling through each rooms, ie. if it takes 15 second to enter and exit the first room, it must also take 15 seconds to enter and exit every other room.

Here's my breakdown of it:

Now, they didn't do it all in one take, of course. They did multiple takes, each one of them with the man in different positions, doing different things. Using a camera motion control it's easy to get the exact same camera movement on each take, which will remove the need for any kind of 3D tracking because you can superimpose each take and they will line up perfectly.

After all the takes are done, it is just a matter of putting it all together in post-production, masking everything out, layering every take's footage and finessing everything to the point where it looks as good as it did on the music video (I made it sound much easier than it actually is).

There it is.
I hope a making of will be put out soon! I wonder, how wrong am I ?