To answer my question, I created a computational poem from my journal entries with Python and Markov chains in a Jupyter Notebook. For the predictive text piece, I transcribed my journal entries around my birthday for the following ages: 15, 16, 17, 19, 21, 22, and 23.
In the output, the majority of the lines are altered snippets from my entries or entirely new. The piece centers on slipping into a new age. I chose the word “slipped” because we do not choose to grow older, we fall into the number. The diction attains a high level of movement.
After generating my output, I I chose to replace one line in each stanza with real images from my journal entries. I realized that between ages 15 to 23, my handwriting became more mature. I found the poem to be largely nonsensical, but my personality definitely ripples underneath.
Code: Journal Entry Generator
I had another question: How do we humanize computer-generated poetry?
In my next iteration, I produced a visual representation of my computational poem in the form of a virtual production. I utilized cinematics in Unreal Engine. I developed a nostalgic tone by constructing a film camera effect in post-processing with Adobe Premiere Pro.
Each character contains an image from my respective age that I manipulated with FaceApp, an AI tool. The final character contains a face made with machine learning in RunwayML.
The setting is a liminal NYC subway car, moving from one location to the next. The transparent characters and setting are inspired by the train scene in Spirited Away by Hayao Miyazaki.