This is a project I completed for NaNoGenMo 2018. As the title suggests, I forced an AI -- specifically the Computer Vision tool from Microsoft's Cognitive Services -- to view and analyze 14,635 frames of the movie Santa Claus Conquers the Martians.
I added some syntax between captions to help it flow better and to convey the AI's lack of confidence about what it was seeing final novel comes in at 64,566 words, which satisfies the requirements.
There's really not much code here. The ek.py script arranged the interface to Microsoft and saved all the image descriptions into a CSV. Then write.py read and arranged the resulting descriptions into something like a novel, using Dominate to generate an HTML file which pdfkit turns into a PDF.
Enjoy!