I’m excited about the upcoming release of the sci-fi survival adventure Transcendent. Right now, you can pre-order your digital copy on Amazon. Paperbacks will be available also, probably a few days ahead of the digital release, just to be sure I’ve got all my ducks in a row.
As a way to promote the book, I’ve been tooling around a lot with an AI image creator called Midjourney. You create a verbal prompt of what sort of image you want and let the AI do the rest.
In my month of testing it, I find that it’s very good at rendering faces but terrible at understanding description of anything more complex than simple nouns.
The AI can’t seem to decode certain verbs or phrases or directives at all (try telling it that a character wears a jacket but no tie—it just doesn’t get it), so you have to keep revising and simplifying how you phrase certain directions and give up at other times completely after numerous attempts when it just doesn’t get it.
As an example, here’s the closest Midjourney came to rendering an image of the Transcendent, the stricken spaceship of the book’s title.
Here was the description I entered into the Midjourney prompt box:
“An interstellar spaceship in deep space. Stars in the background. The spaceship resembles a long tube hundreds of feet long. Attached to the middle of the tube by long spokes is a large revolving ring.”
After 22 attempts:
It’s not bad. (Where are the connecting spokes, though? Is the ring just hovering there by some kind of anti-gravity around the ship?) Still, it’s dramatic, eye-catching, and not to far off from the sketch I made of the ship:
And not too far off from the image of the ship on the cover of the book:
Here’s an image of Troy Hartman that I think Midjourney approximated really well.
Here was my instruction:
“An African-American 38-year-old crew member of an interstellar starship, he looks like Donald Glover, a stern expression as he is preparing to battle a crisis, wears his jumpsuit, head is exposed, close-up, rough-shaven.”
I was impressed with the results:
So, Midjourney knows what Donald Glover looks like, but it was less successful with the next one.
No matter how many times I told the AI that this character has SQUARE glasses, the AI simply didn’t understand. That’s when I wondered if the AI was generating images based on an existing repository of images it could extrapolate from:
Still, the rendering captures the look and spirit of the character. By that way, that’s Franklin Dix above—the federal agent on the trail of the book’s heroes in Edge of Light.
All in all, Midjourney is a major help if you’re an author trying to find ways to render images of your characters and settings. Again, from Edge of Light, here’s a pic of one of the book’s settings:
It’s not exactly what I had in mind, but it captures the windswept and barren peak, complete with the tattered flag, that evokes a key setting from the book.
Keep in mind that these images were the products of dozens and dozens of attempts before the AI rendered what I looking for.
Still, as a promotional tool and based on my own experience with Midjourney, AI image generators are indispensable. They’re not terribly “bright” in how they decipher instruction, but, so far, it’s been a fun software to use and, as an author, it’s been thrilling to see the faces and places that have lived in your head for years given new life as an AI generated image.
Of course, you can also order your copy of Edge of Light too. My audio of chapter six should be in our inbox shortly. Catch my audio reading from the beginning if you haven’t started yet.
Hope you’re all having a wonderful Memorial Day Weekend! Onward and Godspeed.