I’ve been out of action on this site for a while, because real life intruded… And I’ve also been busy writing an AI-powered Android app that generates children’s stories.
As part of the app, I am illustrating the stories with AI-generated images based on AI-composed stories (and using AI to write most of the programming code). I’m primarily using Dall-E, an image generator produced by OpenAI (the creators of ChatGPT), but the pictures, to be frank, are not fantastic – at least, not in comparison to the art I’ve been getting from Midjourney, the better known, more skilled art generator in the current marketplace.
So why am I using Dall-E? Well, there is a programming interface for Dall-E, but not for Midjourney, so I can generate Dall-E art in realtime within the app, something I can’t easily do with Midjourney. Hopefully that will change soon. Either Dall-E will lift its game in terms of image quality, or Midjourney will add a programming interface. Either way, AI generated art will soon be everywhere, embedded into apps and taken for granted.
Here are some examples of the images produced with the two image generators, using word-for-word identical prompts. In each pair of images, the first image is by Dall-E, and the second is by Midjourney.
The images are based on stories that are based of selected images from a random stream of images. The three key words have been chosen semi-randomly. Each image generator had the first paragraph of the story, for context, in addition to the three keywords.
Goblin’s Moonlit Voyage
“A story involving a a goblin… a river… and a mountain”
Wombat’s Watery Way
“A story involving a road… a waterfall… and a wombat.”
Sprite Circus Showdown
“A story involving an evil sprite… a robot… and a circus tent”
Froggy’s Dragon Lesson
“A story involving a frog… a dragon… and a classroom”
Skyward Whiskers Unleashed
“A story involving a plane… a cat… and a furry monster”
“A story involving a stone wall… a monkey… and a pig”