Reposted from Chris Zhou’s blog (Fall 2023), in response to the Visualizations assignment for the Mind Games in Film: Narrative Complexity in Transnational Cinema advanced seminar a few years ago. This was one of the first assignments I received that was (transparently) AI-assisted, and a perfect example of critical and creative co-creation with generative tools.
So what I decided to do is to capitalize on AI platforms’ capabilities to use complex text-prompt to help generate the best visualization as I can. And the prompt I give initially is:
“Create a surreal, dream-like visual collage that merges elements from iconic mind game films. In the center, depict a large, complex maze representing the twisted plot of ‘Oldboy,’ with its paths leading to various scenes from the movie. Around this maze, incorporate fragmented, mirror-like pieces reflecting key scenes from ‘Mulholland Drive,’ showing the blurred line between dream and reality. On one side of the maze, add a montage of memories being selectively erased or altered as in ‘Eternal Sunshine of the Spotless Mind,’ highlighting the theme of memory manipulation.
Adjacent to this, include a series of interconnected, looping time portals, akin to ‘Source Code’ and ‘Timecrimes,’ symbolizing the recurring, cyclical nature of time and events. Each portal should give a glimpse into different pivotal moments from these films, emphasizing the concept of altering past events. In the background, weave in an abstract, Escher-like stairway structure, inspired by ‘The Incident,’ to represent the infinite and paradoxical nature of time and space.
Finally, overlay this collage with translucent, ethereal layers that resemble the interconnected origami from ‘Upstream Color,’ signifying the interconnectedness of all these narratives. The entire visualization should evoke a sense of being lost in a labyrinth of time, memory, and identity, mirroring the complex narrative structures of these films.”
That, obviously, are combining more than five films in our class and as I try to incorporate different themes, I want to see different result. This is my output.
Obviously I tried very hard to incorproate so much for the AI to generate the result I want. I believe it did very well in evoke a sense of being lost in a labyrinth of time, memory, and identity, mirroring the complex narrative structures of these films but they failed to use concrete examples from the film, be it positioning or physical signals.
And then I decide to reduce the film to 2, guess what two films are involved in here: Keep in mind elements are combined in a weird way( so that it’s not easy to tell, and that’s the fun part of it):
Finally, I decide to just focus on one film( which I believe this time is super evident), and I’m sharing my prompt here – I think it’s written pretty well:
“Create a dynamic and layered image representing the key scenes and characters from ‘Rashomon.’ In the foreground, depict the central event of the film: the encounter in the forest. This should be shown from multiple angles, each representing the differing accounts of the characters – the bandit Tajomaru, the samurai, and the samurai’s wife. Each perspective should slightly overlap, symbolizing the conflicting narratives.
Above this scene, in a more ethereal layer, show the rain-soaked Rashomon gate, where the woodcutter, the priest, and the commoner discuss the events, reflecting the theme of subjective truth and moral ambiguity. The expressions of these characters should convey their confusion and disillusionment.
Integrate subtle, ghostly images of the samurai’s spirit as described in the medium’s testimony, adding an element of the supernatural and further complicating the narrative layers.”
What I made differnlty is that instead of keeping it as real as possible, I deliberately add supernatural elements into the prompt so that the output is not the exact impression visually you will have on Rashomon. But it encapsulates its meaning, doesn’t it? …
I think these versions capture the film’s haunting atmosphere and its central theme of the elusive nature of truth, as well as the moral complexities presented by each character’s version of the story.
Expressive AI 