There is something unsettling about the framing in the movie The Boss of It All (Lars von Trier, 2006). Often times the scene compositions seem to disregard its content – the actors, their movements and actions, the objects and settings, everything seems to blend, as if the camera represented an alien perspective, definitely not human.
The reason for this particular perspective is the “cinematographer” built specially for this film: the computational system Automavision.
The way Automavision works is relatively simple. For each scene, first the camera is placed in a default position, more or less faced to the focus of attention. Then, Automavision is activated. The system repositions the camera in an arbitrary fashion, modifying parameters such as pan, tilt and zoom. The first take is then filmed. This is repeated for every following take. (1)
The images in this post illustrate the result – each pair of images represent different takes from the same scene (click on the images to enlarge).
[N]early every cut feels like an ellipsis. A film, von Trier has said, should be as irritating as a pebble in your shoe, and his abrasive tempo gives his comedy an anxious edge. (Another pebble in your shoe)
[Automavision is a] principle for shooting film developed with the intention of limiting human influence by inviting chance in from the cold (…) and thus giving the work an idea-less surface free of the force of habit and aesthetics. (source: press release)
Here the director elaborates further:
Basically I make the frame how I’d like it to be in the film, and then we push this button on the computer and we get given six or eight randomised set-ups – a little tilt, or a movement, or if you should zoom in. It’s supposed to make the image imprecise. And then you say why?(…) I am a man who likes to control things, and if I can’t control them totally I will not control them at all. After doing Europa with very very fixed shots and camera movements, I was tempted to do something totally different. I started using a handheld camera and we invented a form of framing, or non-framing, called pointing of the camera, because I hate framing.
I think it changes the way you watch films because you have to be much more active. Also I am crazy about time cuts. I have a theory that the audience tie everything together so they don’t see time cuts but the time cuts give us the possibility of jumping in time, which means a psychological evolution can be cut down. (Slave to cinema)
Researcher Tim J. Smith makes an interesting analysis in his blog:
However, the natural instinct for framing is hard to overcome and it appears that Von Trier realized that his desire for completely un-framed shots would not be possible so long as he or his camera operator were controlling the camera. (…) If Von Trier is to create shots in which viewers are unable to predict what is the most significant part of a scene or how the camera is going to move he needs to take the human camera operator out of the equation. Hence, Automavision.
In a Von Trier film, as already discussed, the intention appears to be to encourage active search of the screen and disagreement between where different viewers look. With the addition of Automavision the likelihood that a viewer looks at a part of the scene that would traditionally be insignificant probably increase. (It Lookey like Lars Von Trier is at it again.)
The following quote suggests the possibility of “procedural directors” in digital narrative experiences, like in games:
As suggested by Mr. Smith’s closing comment “Anybody up for a game of Halo directed by Lars Von Trier?”, Automavision hints at a future time where cameras in-game aren’t just patterned after object recognition and collision detection, but also take into consideration more artistic merits, such as emotional impact and more ‘classical compositional’ attributes, as well as being able to mathematically deal with more auteuristic visual narrative elements. To re-use the film director analogy, imagine having the choice between having say, God of War directed by Von Trier (an incongruent and edit-happy, but still somewhat understandable, mess) or say, Spielberg with loads of low-angle pseudo-tatami shots? And why stop with directors, why not cinematographers? Roger Deakins has quite the eye, and maybe we could programmatically raise the ghost of Sacha Vierny. (Automavision & Lookey) (3)
Note that, even though there is a strong experimental quality to the use of a system like Automavision, the human presence is still very much there. Lars decides which takes to keep.
One afternoon, everyone waited around for three hours, because the program kept pointing the camera at a blank wall. “Lars was furious with himself,” recalls the actress Iben Hjejle. “He said, ‘Why did I come up with this idea, the stupidest idea I’ve ever had?’
(Lars Von Trier’s funny turn)
Regardless of the intentions and meanings behind this particular technique, it illustrates a procedural creative strategy, with poetic (or expressive) potential.
[This is an abbreviated version of the original post in portuguese.]
1. There are many different references about the Automavision system. I based this post in the movie’s press release, as well as interviews and articles. (voltar)
2. This approach of loss of control is similar to that of the Dogme 95 movement, co-created by Lars in 1995. Interestingly, while in Dogme the simplicity is obtained by removing technology, in The Boss of It All framing is disregarded through the very use of a computational system. (voltar)
4. Another interesting aspect of the use of Automavision is the way it affects the actors. “[O]ne of the main reasons for using Automavision was to ensure the actors couldn’t use any of their usual tricks. Thanks to the randomised framing and audio settings, they had no idea of how the camera was going to behave and therefore weren’t able to try to show off their best side or steal scenes. (Von Trier’s original idea had been to hide the camera altogether and film through a double mirror, “but we had too little light. We couldn’t do it”.)” (source) (voltar)See more researchblog.