Generating images from text prompts is here, and it’s going to empower all writers who can see pictures in their head and have the power to describe them with words.
This is a big deal for anyone who isn’t visually creative, but has lots of ideas.
Ai Art is here and ready to use by anyone* The astrix being it’s a little unfriendly to face right now, but those who are happy to follow a few steps, will be able to generate incredible illustrations, 3D designs, and art with full painterly results like the masters from that Davinci bloke to Banksy.
The two images above were created in about 15mins each by software following a one sentence description. The images are very impressive and could have taken a traditional 3d artist or digital illustrator weeks of time effort and cost.
Look closer at the images. What can you see exactly? You can see (on the left) a sky, some kind of buildings, some rock type structures in the lower half. On the right, the image is a landscape view of people, and rock formations, and in the distance there’s more people possibly, and light in the background. Maybe a way out of the cavernous landscape?
These two impressive images aren’t actually anything specific. They are a suggestion, a feel, a mood, a vague theme. You get a feeling, a vibe. You don’t get detail, specific detail. Yet.
This is for two reasons: the quality of the input text, and secondly, the patience needed to generate image after image until the desired result is reached. It is definitely not a one and done solution, yet.
But us fiction writers are patient! And we do need images of ‘feeling’ and ‘mood’. Think about contacting a designer or illustrator and trying to explain what you want your book cover to feel like. Ai Art could be your first challenge easily vaulted.
This next image is of an explosion in a city street with a painterly, watercolour type effect.
Again, there is a suggestion a lot of things but the more you look the more vague the image is.
These examples make for perfect book covers: all theme, feeling, vibe, and suggestion. Ready for Text to be laid over the top and presented to the world.
The images above are from the following text prompt:
‘A photorealistic futuristic city at night with vehicles hovering overhead casting spotlights on pedestrians by Greg Rutkowski, rendered by unreal engine’
The gallery of 6 images show the stages of Ai art generation: 17%, 34%, 50%, 67%, 83%, & 100%.
The renders are from top left (the first), showing a vague splattering of light and shade. The second (top middle) image shows much more detail, but the software still doesn’t have a final image in mind yet. It’s just seeing where the image takes it based on the input prompt. The third (top right) image show much more cohesive shape and form of a night city scene. The rest of the images (bottom left, centre, right) are just building in the details. Click the images to see closer detail.
This attempt was 1 of 1. What other results could be generated from a prompt if I let it run many times?
The above images were created from:
‘A beautiful painting of a singular lighthouse, shining its light across a tumultuous sea by greg rutkowski and thomas kinkade, Trending on artstation.’ & ‘yellow color scheme’
The Greg, Thomas, and Artstation references are simply a trick to get the Ai to paint in a specific style and quality. I’ve displayed many of the stages, and below are three more results from this same prompt.
Want to know how to do this yourself?
It’s time to Google or search on Youtube for Disco Diffusion Ai Art. You’ll learn a lot, and end up on https://discodiffusion.com/ and probably reading about Zippy’s Disco Diffusion Cheatsheet
There’s so much to learn, and we’re all just getting started. In time as the software improves and more apps are developed we’ll have specific book cover creation software. But until then you’ll have a hell of a time playing with these great free tools and creating artwork you can only dream of.
Good luck, and let me know how you get on!