TECHNES Encyclopedia Database

Signal Manipulation and Datamoshing, by Agathe Presselin

A third kind of intervention concerns the manipulation, not of the film stock, but of the digital signal. Datamoshing, part of a larger body of practices known as glitch art, plays on the way a digital image, moving or fixed, is read, and has been a popular technique amongst re-use video makers since the mid-aughts.

To understand datamoshing, one must go back to the principle of the digital image and of compression. From both an economic and a practical perspective, the digital image industry made two main arguments: the quality of the image and the ease with which they can be stored. In order for these two qualities to be in harmony – no easy feat, because the better the image quality, the larger the file – the file is compressed, meaning that its data is reduced, sometimes radically, without harming its legibility once it appears on a screen. In compression, a sequence of images is distributed amongst several GOP (groups of pictures), which define the organization and legibility of the shots. Each group of images contains an I frame and a P frame. The former is the key image, or keyframe; it is always found at the beginning of the group and contains all the data describing the graphic elements of the image (the decor, characters, etc.). In other words, all its pixels are coded. The P frames are all the following images (which are mixed in with the B frames), but these contain only descriptions which relate to those pixels pertaining to effecting movement. When one shot ends and yields to a new shot, one group of pictures gives way to a new group of pictures. Datamoshing consists in corrupting the data in a digital video in order to suppress the I frames, or to invert the different kinds of frames. In this way, the “correct” linking of shots – of the distinct units which follow one after the other to narrate the action, for example – is disrupted, and the successive shots merge into a heterogeneous mass of pixels.

The film Résidus (2015), by Laurie Mannessier and Pierre Simion, uses datamoshing to mix up digital images of every description (music videos, films, 2D and 3D animated films) and create singular and original visual linkages between them. Each re-used shot is of short duration, rarely being in motion for more than three seconds before freezing under the effect of data compression. And yet its presence on screen exceeds the duration of this movement, as the frozen image remains visible until it is entirely dissolved and covered up by another. What is interesting about the use of datamoshing in this film is essentially that it prompts us to rethink, under the digital regime, the ways in which shots are linked.

Take, for example, a shot from the beginning of the film showing Homer Simpson hanging from a wrecking ball in the middle of a construction site. The first image is the I frame: Homer, the ball, the decor. The swinging of Homer and the ball that follows is divided into P frames, each tied to the reference image (the I frame) and containing data describing the relative differences from this initial image (the ball having moved to the left, Homer begins to turn his head to the left). This initial series of images is followed by another film clip, this time a live action sequence, showing men arguing on a lawn. The image of two men standing and a third on all fours constitutes the I frame, while the rest of the shot (the man on the ground approaches) is a group of P frames. By getting rid of the I frame, Homer Simpson is shown frozen and “under attack” by the group of men in the following shot.

The P frames are clearly connected to the I frame which precedes and, in a certain sense, pre-defines our understanding of them. But if the I frame is done away with, the movement data in the P frames will play out with respect to the previous I frame (on condition of using software which does not balk at applying what it is supposed to treat as anomalies). The reason the men appear as if cast in the picture of Homer is because, brutally extracted from their own decor, the I frame corresponding to the beginning of this group of images has been removed. All that remains are the moving pixels which make up the three companions. It is this aesthetic and technical situation which the few words uttered by one of these men, conveyed by a subtitle floating in this pixel soup, ironically – and involuntarily – conveys: “This is not natural.”