Counterfacts I - Zapruder29/09/2025
Model: Veo 3 Fast
1280x720
24fps
01:08:57
(517 shots of 192 frames)
Frame 312 of the Zapruder Film (the frame immediately preceding the moment of impact) is used as the starting point for a stream of variations, using the image to video model Veo3. The variations are produced by using as input the same frame every time, along with an extremely succinct textual description relating to the possibility or impossibility space of the elements in the frame, like: “The car thinks”, or “The people normalize”. These produce at times subtle variations in the trajectory indicated by the image itself, or causes totally unexpected events in the subsequent 8 seconds, after which the loop resets.

One of the most dissected, analyzed images in film history, one in which has been sought the truth of what really happened, is declensed into a portion of the set of events that didn’t, in an impossible attempt to exhaust the space of possibility.

One could also imagine:

• The film that takes as its starting point every single frame of the Zapruder film, creating a film composed of exactly 486 shots that each show one version of what didn’t happen for the subsequent 8 seconds.
• The film that takes as its starting point the first frame of the Zapruder film and extends it endlessly, reusing the last frame of each generation as the new starting point.
• The film that, using each or one single frame of the Zapruder film as a starting point, attempts through precise textual description to approach the image of what really happened.
  
• The film that generates from Frame 312 backward in time, creating a reverse causality where the "end" generates its own fictional origins.
• The film that takes frames at regular intervals (every 10th frame) and generates forward exactly until the timestamp of the next source frame, creating a parallel timeline that keeps touching the original.
• The film that generates from multiple frames simultaneously, creating split-screen impossibilities that diverge from the same historical moment.
  
• The film that uses only prompts describing what is not visible in each frame ("the grassy knoll", "the second shooter", "the missing frames") - generating the conspiracy theories themselves.
• The film that uses forensic/technical language only ("trajectory adjusts 15 degrees", "velocity decreases", "temporal dilation occurs") - treating physics itself as the variable.
• The film that generates N variations from each frame with identical prompts, showing the model's own uncertainty/interpretation space.
• The film that nests generations: generate from 312, take a frame from that output, generate from it, repeating until the connection to the source becomes unrecognizable.

And so on.

Copyright / Guillaume Menguy / 2025