MEMORIA is an immersive four-channel expanded cinema installation with specifically developed software and hardware components. It combines ethnographic research, science fiction, and “recombinatory” forms of cinema that appropriate found or pirated footage. The work intertwines two narrative planes: the dystopian vision of William Gibson’s Johnny Mnemonic (1981), one of the earliest cyberpunk short stories, and the real-world Cuban “offline internet”, the physical data distribution network Paquete Semanal. The work takes the form of a “documentary remake” in which actors, performers, and actual Paqueteros embody Gibson’s fictional characters. Havana stands in for the “Sprawl”, the urban dystopia Gibson describes as a ruin of the future where technological and scientific advances, such as artificial intelligence and cybernetics go hand in hand with the breakdown of the social order.
MEMORIA’s narrative is presented on 4 synchronized screens that merge live-action footage with found footage from videos that have circulated in Cuba through the Paquete. These materials come from ARCA (Archive), an artwork consisting of a 64 TB server that contains the entire content of Paquete Semanal for an entire year (52 weekly issues). It remains the only large-scale archive of this ephemeral phenomenon.
Installation with 4 video channels, a central server, adapted hardware and software, and computer chairs.
The installation surrounds the viewer with 4 synchronized screens of variable sizes, hanging from the ceiling and transparent so that the projected image can be seen from both sides. Between the screens are several rotating computer chairs that invite the audience to sit.
To represent the protagonist's loss of memory, the work will slowly "die" over the course of the exhibition. For this purpose, we have developed a special hardware setup consisting of a computer with a video card with four 4K outputs, which acts as the central server of the installation. From this server, the four video channels and a 4.1 audio track are played back synchronously. This computer also runs Lost.CU a bash script that executes various programs that continuously manipulate the video: FFMPG, a powerful free software for video and audio transcoding, and two custom programs written in Ruby: Datamosh, which deletes segments of the video so they can be overwritten afterwards, and Moshy, which overwrites these empty segments with information copied from random other parts of the video. This process gives the impression that the visual information in the videos is gradually lost and the image progressively degrades. However, the process does just the opposite. The video files are enriched with their own information, so the file increases in size during this copy and overwrite process. Since there is no linear header in the file to convert these bits into pixels, they become visible as artifacts that affect the quality and smoothness of the videos. This automated, constant copying process not only corrupts the audiovisual material, but eventually also damages the hard disk where the project files are stored. The speed at which the material deteriorates is adjusted to the duration of the work's exhibition. The copy script basically operates like a tumor virus, like a cancer that makes the cells inside the host grow in a random and uncontrolled way. Every day it produces a new set of video files that amplifies the changes of the previous set: what appear to be subtle changes in the first generations of video files become more and more obvious in the later generations, eventually causing the information to become completely distorted over time as the video devours itself through the constant process of faulty replication. This process corresponds to what happens to Johnny in the narrative when he loses his own memories due to the overload of information in his brain.
The Memoria project is implemented using Python and the VLC media player. The main components of the project include:
This function takes an input video file and applies a noise filter to it using FFmpeg. The corrupted video is saved as a temporary file with a new name specified by the output_file_template parameter. The original input file is then deleted, and the temporary file is renamed to the original file name.
This function plays a given video file on the VLC media player, using the specified screen nuasdasdasd asdasd asdasd asd This function applies the corrupt_video() function to a set of input video files, creating a set of corrupted video files that will be played by the play_video() function.
This function starts a set of VLC media player instances, each playing one of the given video files using a unique screen number. The function returns a list of subprocess.Popen objects representing the running VLC processes.
This function stops a set of running VLC media player instances by killing their corresponding subprocess.Popen objects.
The main script for the project is memoria.py, which sets up the schedule for periodically corrupting and restarting the videos. The schedule is created using the schedule module, and the scheduled_task() function is called at a set time each day to stop the current videos, process new corrupted videos, and start playing them.
To use the Memoria project, you'll need to have Python and the VLC media player installed on your system. Clone the project repository, navigate to the project directory, and run the following command:
python3 final.py
When the project was set up in Aksioma, HDMI audio interfaces failed to appear. Seems it was a bug in Ubuntu 22.10 (or was it 23.04?) default kernel. Upgrading to the latest 6.1 kernel fixed that. There's a handy script here. Usually running the latest kernel is not the best choice but in this case, bugs were fixed.
Audio interfaces appeared but speakers did not match the projections shown. So we switched VLC to JACK audio and modified the script to include the video IDs in audio output names. Then used qpwgraph to match audio outputs to projections. Connections can be persisted so future glitched videos still match outputs.
KopeK
g1smo for all the help with the audio, thanks m8 !