After my cousin Roisin told me about the AI Generated images for “salmon in the river“, I thought I would share some of my own AI image fails. Working with AI generators is still a process, still needs the human hand, and it often takes a great deal of thinking and adjusting to create the kind of images that you want. Sometimes the AI just cannot get what you are feeding it. I’m illustrating my next book of short stories with images that are, in part, being created using the AI image generator Midjourney. In one of the stories, “Nature”, a toy poodle named Puffy runs along the back of the couch, barking at everything he sees out the window. Here are multiple #midjourneyfails that have happened while trying to create the basis for this image. Midjourney doesn’t seem to be able to totally understand “poodle”, let alone the idea of a poodle on a couch. I’m still trying to create a workable image for Puffy, but I’m getting there.

I am using text from the stories as prompts. These are results from the prompt: “The dog is a white toy poodle. He spends much of his time standing on the back of the couch, yapping at whatever he sees out the living room window.”

And here are a few images from the prompt “A white poodle running down a forest trail”: