Edit: this was done in 2015, when generative ML wasn't amazing. I also had zero experience with ML.
I really wanted to try out the neural network producing Trump tweets
The are tutorials online on how to generate a text from a set of words.
I decided to go with romantic texts because a decent database was available easily.
After a while, I didn't really know what to input as text set. At the same time, I wanted to try out Image captioning with TensorFlow from Google
I downloaded my gf's instagram posts, mixed them with random free images with #nature found online and launched the bot. These images were fed to the caption extraction model. The output set of words was piped to the romantic short story writer model. Eventually, I glued it on a Django blog framework called Wagtail, generating posts on demand.
The tags of the blog posts are the captions extracted from the image.
This post was written more than a year after I completed this project
After a year and a half, the results in page views are not great. 6ven.com has 617 pages and yet it gets only around 50 hits per month.
Original full portrait: "Playmate of the Month". Playboy Magazine. November 1972, photographed by Dwight Hooker.This 512x512 electronic/mechanical scan of a section of the full portrait: Alexander Sawchuk and two others - The USC-SIPI image database, Fair use, Link
.png){kind=link}