Thoughts?
This is for an assignment due in 2 weeks. My daughter had surgery last week and I wasn't sure how much time I would have post-surgery so I worked ahead. I also invested way more time in my last two projects (5 watercolors bound into a cut-out flip book for the first one and a quilted appropriation of "The Scream" for the second), so I wanted to give myself a bit of a break on this one.
The assignment is narrative extrapolation: "Using the information gathered from (a) story, design a creative piece that
demonstrates or explains the central themes of the story." We're allowed pretty much free reign as to media.
The short story I chose was Ursula LeGuin's "The Ones Who Walk Away From Omelas" https://shsdavisapes.pbworks.com/f/Omelas.pdf TLDR: it is essentially The Picture of Dorian Gray - a bargain with the devil, but on a community scale (with a small child instead of a portrait in the attic paying the price for perfection in the rest of the community). Those who walk away are those who have decided the bargain isn't worth it.
Here's the video (images with transitions) I created for my project: https://vimeo.com/1077827036/3c457c37d2?share=copy
(Sorry - I tried to upload it to Imgur, but Imgur wasn't cooperating so I had to shift to vimeo
I used Adobe's Artificial Intelligence tool Firefly to create two starting images. Firefly was trained solely on licensed images - which eliminates most of my ethical concerns about using AI. I used those two images to create 4 additional images, modified in Photoshop to match three themes in the story (Utopia, the sacrifice of the child, the shattering of Utopia should anyone dare be kind to the child) and an ending that has at least two interpretations.
Firefly seems pretty clunky to me, from descriptions I've seen of using other AI image generators, but I didn't want to use AI that stole others' art so I didn't even test alternatives. I spent 6-8 hours playing with text descriptions to generate an image that I liked, including playing with uploading some of my own photos as a starting point (that didn't work well), trying out different themes, etc. I generated about 1000 total images/alterations of generated images to get the one I used. The thing that surprised me was how hard it was to get Firefly to follow relatively simple directions. The child in the final image is "painted in" to the first image I used. Out of maybe 100 images "painted in" to that location, I could get only a handful with the child facing away from the buildings. (I tried walking away, facing the foreground, face visible, crying, smiling (thinking the last two would force it to generate a child facing forward in order to show crying or smiling)).