This part is going to be a demonstration of image development in MJ. I can hear your thoughts hammering through the universal consciousness, this is AI for Writers, why should I be interested in generating images with AI?

I'm glad you asked. A lot of artists, and writing is an art form, will talk about visualising the mind's eye. This ability to form images inside the brain without any external input and then project them into the real world via an artistic medium. My belief is that it's great if you have it. Congratulations. 

My mind's eye is somewhat shielded by a dense cataract and, as hard as I try and look through the opaque cerebral tissue, the image is blurred--I know it is there, but what exactly? Is it Castle Erlend that I'm trying to picture; a mutated king rat that is about to attack; the close, detailed face of a lover (who is about to be attacked by King Rat)? 

Is the image in your mind's eye clear and lucid? Could you pick up a pen and transpose the shapes and details; shades of light directly on to a piece of paper? If so would it be a well defined as the image below?

BTW, the image above is a two minute sketch of King Rat with my phone stylus. I am a marginally better drawer than this, but for this exercise it's a pretty good approximation of what I see with my mind's eye when I think about this fantastical beast that may or may not form in nature when a colony of rats develops in a confined space; so confined that their tails grow together and they must move and feed in a single mass.

If I am writing about King Rat, then I want my inspirational vision to be as detailed and interesting as possible so that I can take this image and move it to the page. The better this vision, the more likely we will be able to distil convincing smell and sounds for the scene.

I feed my King Rat master piece into MJ because I wanted to give myself a head start. When the adventurers encounter this beast in the dungeon, I know exactly what I'm describing and I can write it as close or as different to this image, as inspiration dictates. The image generator, I was confident, would give me something worth writing about.

Following is a series of MJ '/imagine' prompts that show the development of King Rat. 

1. At the time of writing this (April, 2023 -- this content will date very quickly) when I suggest the following prompt to MJ v5, the next image is the result:

/imagine: a number of monster rats living in a dungeon sewer in a closed space so that their tails grow together and they appear a big mass of joined rats. dark. dirty. disgusting.

Urrgh, I did warn you.

Following a prompt MJ will give you a set of four options. You can make variations of each image or choose to upscale. While oddly inspiring these images were not a great representation of the image I had in mind. 

2. In the next prompt, I added my sketch to give the AI model something to start with that was closer to what I had in mind.

\imagine http://s.mj.run/nR1S28iwyBQ  number of monster rats joined by the tail. dark. dirty. disgusting

While the top right option look like salamanders, the others are very ratty. It's a start. Don't be disappointed at this point, it gets better, stick with it.

3. I made variations on 3 and 4 (bottom left, bottom right) to see if I could get the tails to join up. After a few more variations, I upscaled the below image. The idea was then to use this, mass of rats, as a new image prompt.

could it get any better that this?!

/imagine [image web address here] a number of raging monster rats joined by the tail. dark. dirty. disgusting

Yes! Already, after only a few iterations, we are a long way from the scribble.

Looking back, I could have used image 1. from here, but I re-rolled the prompt and took a variation on 2. In the end I went with the following image because, to me it looked dynamic and showed the back of one of the animals. This formed the image prompt for my next \imagine : 

4. It was time to work on the text of the prompt. By adding a double colon "::"  followed by a number, you can add weightings to elements of your prompts. To be honest, I'm not sure if the parentheses make any difference, but I like to think it was helpful for grouping elements of an image.

/imagine [image web address here] a photograph of a group of (giant raging monster rats)::1.3 (joined by their tails)::1.4 in the sewers. dark. dirty. disgusting

The next round provided a group of very similar option, all an improvement for the source image. I chose the first:

a, couple of more such iterations, and I started to lose it (as shown below). So I rolled back to the above image and updated the prompt text.

5. In this prompt I introduce some more descriptive items into the prompt text. "foam and slime and red sauce is dripping from their fangs" (MJ is PG13 so a lot of words like 'blood' and 'gore' are prohibited from the prompts). Also added "tails like knotted rope" to try and bring in the King Rat element.

/imagine [image web address here] a photograph of a group of (giant rabid rats) (joined by their tails like knotted rope)::1.4 in the sewers. foam and slime and red sauce is dripping from their fangs. red eyes. dirty. disgusting. diseased

I'm going to skip the commentary for a bit, as the process follow that these steps iteratively until you get something like the final image below:

and then, something interesting happened:

/imagine [image web address here] photograph of a group of angry dog rat hybrids with their tails joined together in a knotted mass of rope. they have sharp fangs::1.3 dripping red sauce. red and green eyes. slimy sewers

I knew I was on a winner when Lorna demanded to know why on earth I would show her something so disturbing. 

My job from here is to bring life on the page to this blind, tic-eyed evolution of King Rat: where does it live; how does it feed and on what? Now, when I'm writing I have an image in my mind's eye to work from, an image that no one has seen before and will, hopefully, be more vivid on the page than if all I had to work from was one lame scribble.

<◕.◕>