AI Art blog article
Over the course of writing this blog, I have often fallen onto using public domain art to illustrate the incoherent thoughts that I tumble onto these pages. As I ventured into short stories, pictures were much harder to come by. I usually needed a very particular image to add to the story, and amidst the billions of images, I was never really happy with what I could find. Unfortunately, commissioning art is out of the question. I write not for fame and fortune, but to document my hobby life and all the things I find interesting, and then toss in a few stories to get them out of my head. So, there isn’t any money in this, least of all any extra money to use to commission art. That said, I have commissioned a few pieces. Just a handful. Four or five, I think. All are brilliant, and the two artists were fantastic to work with and created exactly what I wanted each time. There are many, many artists out there. Prices are all over the place, depending on what you want, how detailed, how many revisions are allowed, and who the artist is.
I choose two artists on separate criteria. The first artist was contacted to create a picture that depicted a stained-glass window. I wanted the piece to use in a terrain piece I have being planning/building/dreaming about for a few years. I wanted a fantasy window depicting a character from Bretonnian lore in the Warhammer Fantasy universe. I knew the artist personally, and saw her create something similar on her own. We set up a contract, agreed on a price, and she delivered this incredible piece of art. When the rest of the Grail Chapel is ready, I will print this on acetate and install it as window, backlighting it with flickering LED “torches”. I think it will be quite spectacular.

The second artist (and honestly I can’t remember where we first ‘met’) was commissioned to create some character sketches for our DnD campaign. The key to these images was price. I needed more than one, so they had to be pretty cheap. I wanted something close to the pen and ink illustrations in the original Red Book for Basic DnD my role-playing adventure started with. In particular, this image by Larry Elmore.

It is a relatively simple image. Nothing flashy or wild. As a younger lad, this image, and the others like it in the Red Book, captured my imagination and lit a fire for gaming that has never diminished.
I was contacted by an artist in Brazil off one of the forums I peruse, and the process started again. We had a discussion, set a contract, preliminary images were shared, revisions were made, and a complete image was delivered. This guy is great. I send him an email with the basic description of the character, he replies right away and delivers a preliminary sketch within a couple days, we hash out any changes and a day or two later I get a complete image. Easy. These are what he has delivered so far. Dynamic, evocative characters right out of our DnD campaign.



In case you are wondering, there is Malark, howling in rage. Seraphina, slyly ready to slice and dice with her twin blades, and Lia, her unique wand of fireballs wrapped around a wrist, ready to deliver fiery retribution.
As cool as these are, they cost money for each image. When I need exact images, this will be the route I pursue. For more images though, I needed something else.
On another forum or thread or Facebook group, I saw some AI images one of the members had created. I was blown away. These images were as good as anything many of the digital content creators were selling as commission pieces. Incredible detail, full color, and a variety of subjects. This was something I had to explore.
My first attempt had mixed, and frustrating results. I tried the MidgardAI. It is utilized off Discord, which is an app I had, but only for following a single Kickstarter campaign, so my experience with Discord was fairly limited. I don’t even recall exactly how you use Midgard. In the open chat you type the prompts with some other command, and in a few minutes you image appears. Also, in the chat thread. You can somehow (I never quite figured that out) upscale and evolve the initial image.
My biggest complaint was that this was happening in a chat thread. Once your image was generated, depending on how busy the thread was, the image was a few to many items away from your initial prompt. When the chat was hot, stuff was appearing every second. It might be mildly annoying, or even perfectly acceptable to some, but I found it pretty irritating. I don’t enjoy chasing stuff on a rapidly populating thread. Weirdly, I did not download the image i created. Or images. I don’t even remember how much i used it before I quit. I think you had one use a day.
What else was out there? With a few searches I next came to the StarryAI program. You can explore other people’s images, and there is some super cool stuff. From Anime to Furries, Fantasy to Goth characters, much of it seems to center on characters. I gave it a shot, throwing some prompts in and getting this as one of my initial images:
I used this simple prompt: Elven Druid Cloak Long Sword Staff Leather Armor

Weirdly, I don’t think of the four images created off that prompt show leather armor. In fact, I think the AI sees Armor and instantly depicts metal. It cannot decipher Chainmail either. Specific limitations, but fairly significant for fantasy characters.
StarryAI has a bunch of settings, and I will not be getting into all of that. The basic prompts and settings generates 4 images for one credit. The above image is one of those four. You can evolve and upscale these images in a much easier manner than with Midgard. Here is one of the first upscaled images I created.

I was trying to get an image for my daughter’s sorcerer, Lia, and the prompt didn’t say anything about a hat. But it created a very cool image.
The only down side to StarryAI is that you only get 5 free credits a day. Surprisingly, once you claim your credits for the day, they don’t seem to expire. On a day I don’t have time to play with the program, I just claim my credits and save them for later. Of course, you can buy more credits, and better access to the AI servers, but I was looking for cheap. Free is the best kind of cheap.
The first images were rough. I spent some time on the Google looking for hints and tips to creating better images. What I really want is the exact prompts some of the best creators are using, alas, none are sharing. I found some helps, though and quickly started creating pretty decent character sketches and landscapes. Some came out very weird. The following image on the right was created using prompts for a female warrior with red hair. I don’t know about you, but I am pretty sure that is a male face, despite the boob-armor. The image on the left is the same prompt, adding ‘human’ to the beginning of the prompt. Definitely a female warrior.


By now, I have created 65+ sets of images using StarryAI. Some have weirdly been so horrible that I deleted them from the My Creations area of the site. Here are some of my favorites:







Quickly depleting my StarryAI credits, I wanted more AI, and discovered DALL-E. To use that site, you have to set an account and log-in. StarryAI seems to be based of your IP maybe. Honestly, I am not sure. I have a profile, but I do not recall how I started on the site. Every time I go to the site my creations are there. DALL-E gives something like 50 credits up front, and 15 a month after. It is supposed to be more plain language friendly. They claim you just type in what you want to create, and the AI works its magic. I have had mixed returns with DALL-E. Some stuff is super cool. Some stuff is weirdly distorted. Faces in particular often seem to be melted wax, or scary eyes. There is less detail and crispness as well. Of course, it might be my prompts and lack of key phrases to lock in specific features.
Here a few of my favorite images created using DALL-E:



Both programs have some weird issues. Both are hit or miss with weapons. Sometimes the weapon hovers in front of the hand, or it starts weirdly near the hand. Many times, the weapons are not fully developed and don’t resemble any sword, mace, spear or staff anyone on Earth would recognize. StarryAI doesn’t understand Single Story Building, either. Any time I say building, it ends up multi-story. I tried DALL-E and get single story, but if the description says lodge, it might generate an outhouse or shed.
I’ve used both for portraits, landscapes, ships, and buildings For character portraits though, StarryAI wins hands down.
So far, I have only used these AI programs for fantasy art. I am curious how a more pulp or historical figure might turn out. Will I be able to use these to add more art to my short stories? Can I get the generator to produce a Wild West gunslinger? What about an Indiana Jones-type Pulp character? I’ve seen sci-fi characters, but can it do a theme like Star Wars or 40K?
It’s a relaxing game, adding a descriptive prompt and waiting to see what the AI generates. So far, I can’t get exactly what I want. I can’t give a list of details, and know they will be in the final image. Which I can absolutely do with an actual artist. Things like jewelry, tattoos, equipment or weapons will be reliably generated with a human artist. I can even say I want silver jewelry with sapphire stones and know I will get that combination. AI is fun, and much cheaper, but not capable of replacing a real artist for quality work. Is it fun to generate images? Sure. Will I have plenty of images to depict NPCs in our campaign? Absolutely. Will I be able to get every image I every need? Nope. There will always be a need for real artists, taking prompts and descriptions and making revisions until my image is realized.
One frustrating image I can’t even get either AI to generate is a single-masted ship. Maybe I need to say sloop or bark? Off to the AI, to see what comes next.