I run a Dungeon & Dragons campaign for four students that meets for two hours a day, two days a week. Our campaign takes place in my home-brew world, Migia.
I don’t have the time or the skills to map a logical geography for a whole world, so I used the open-source Fantasy Map Generator created by GitHub user Azgaar, a JavaScript wizard from Richmond, Virginia, to generate and tweak a world map for Migia.
The generator allows you to customize place names, stylize the design, focus the map on political borders, biomes, cultural zones, religions, etc., and render it as a flat map, a 3D scene, or a globe.
From there, the Internet offers a plethora of D&D-focused generators to help me bring Migia to life. There’s the city map generator, the dungeon generator and cave generator, the random encounter generator, the side quest generator, the backstory generator, the NPC generator, the fantasy name generator, and a whole list of auto-roll tables that will generate everything from a “breakfast at a traveler’s inn” to “resurrection consequences.” On days I prefer rolling dice to pressing buttons, I head over to the D100 tables on DNDSPEAK.com for inspiration.
Generators have been around as long as D&D has. Computers may make them easier and faster to use by combining a slew of tables into one button push, but at bottom, most generators are just the lists of tables Dungeon Masters find in the official rule books or create on their own.
In other words, generators are not artificial intelligence but glorified spreadsheets.
Midjourney Bot To Create Images For A Campaign
In September, one of my D&D buddies added the Midjourney Bot to our Discord chat server. You enter a few words in the chat, and the bot will create an original image based on your prompt using artificial intelligence and machine learning.
I started using it to create images of non-player characters in the campaign.
Take the image of the harengon, for example (a harengon is a kind of rabbit-like creature). I prompted the bot to create “a ferocious rabbit standing on the edge of a cliff with a sword in her hand.” After about a minute, Midjourney Bot provided me with four drafts based on the prompt.
From there, I selected the draft in the bottom right and told the bot to “upscale” it. A minute later, I had the final image of my harengon. The upscale added brush strokes and more detail across the entire canvas. I could have continued to tweak, but the image served my needs, so boom…two minutes after having the idea for a ferocious rabbit sword fighter, I had a picture I could base the doe on.
The results of my various prompts were interesting and captured the vibes I intended (e.g., “a female elf with long white hair and dark skin standing in the center of a cloud made from daggers”), but they lacked the details that I need to really dig into their characters.
Then I realized that I could use the word “portrait” in my prompts to force the bot to create more realistic images of my NPCs.
For the last few months, whenever I’ve needed an NPC for the campaign, I type a few words into Midjourney Bot that summarize the kind of character I’m looking for and add the word “portrait.” A couple of minutes and a few drafts later, I’ve got an image I can show my players to help them imagine the individuals they encountered during the campaign.
From Characters To Scenes
A few weeks after I started using Midjourney, I decided I could use it to create illustrations of the previous D&D session’s scenes.
Most D&D campaigns last weeks, months, years, and sometimes decades. When you sit down with your friends (or students) for a session, there’s usually a few minutes of recap (“Last time on Dungeons & Dragons…”) to remind everyone where we left off.
How much cooler would those recaps be if I could throw in an illustration or two of where they were or the monsters they were still facing?
Thanks to the Midjourney artificial intelligence image creator, my D&D campaigns became much more visual.
Craft Assistant (GPT-3) To Write Original Histories & Backstories
I use Craft to manage all the information I need for Migia. Craft is like a note-taking app on steroids. I discovered it a few months back (long after Apple named it the “2021 Mac App of the Year”), and it’s the first app in a long time that I loved using. I immediately converted 90% of my document-creating/managing tasks to Craft; months later, I haven’t looked back (I really should write a blog post about it; it’s so good).
A couple of weeks ago, the folks behind Craft added GPT-3 to the app, calling the feature the “Craft AI Assistant.” As Craft wrote in their announcement, “We believe that GPT-3, one of the most impressive AI systems ever built, which applies machine learning to understand questions and generate human-like text, has now reached the point where it’s more than just a novelty.”
I tried it out — “Generate a list of blog post ideas” — but didn’t really see a way to integrate it into my daily habits, so I moved on with my life.
But on Wednesday this week, I found myself behind the 8-ball for this week’s D&D session with my students. The adventurers were on a ship on their way to a pirate haven named Maroon Cave, about 200 miles off the coast. I knew a giant shark was about to attack them (I’d ended the last session by playing the theme to JAWS), but if they made it past the shark, I didn’t know what they’d find at the pirate haven.
With less than an hour to prep, I opened Craft, brought up the Assistant, and asked it to “write a 400 word history of a pirate haven named Maroon Cave.”
Within seconds, the Craft AI Assistant had written an entire history, explaining why it was named Maroon Cave (from the French word for “fugitive”) and why pirates were safe there. It also attached famous pirates such as Blackbeard to the history, provided details on two specific events that occurred there, and summarized what the cave is used for today (“a popular tourist attraction”).
I couldn’t use all of that in my campaign, but after a few tweaks (French became Elvish, for example, and Blackbeard became a goblin pirate queen; plus, I deleted the section on it becoming a tourist spot), it was perfect.
Next, I needed a popular location at the pirate haven, a tavern where the adventurers could go to find the information that brought them to the island in the first place.
After a quick stop at the tavern name generator, tweaked with the help of some students who were in the room with me, I asked the Craft AI Assistant to “write a biography of a dwarven woman who runs a tavern named the Tipsy Turtle on an island that serves as a pirate haven.”
The Assistant created a story that begins, “Lydia, or ‘Lyd’, is a dwarf who runs a tavern, the Tipsy Turtle, on Maroon Cave. The cave is ruled by the law of the sea – piracy. Lyd, however, brings a touch of stability to this dangerous place.”
First, note how it recognized I wanted the tavern to be on Maroon Cave. I didn’t tell it that. It picked up the information from the other notes in my app.
It continued the story, “She arrived at the cave roughly 15 years ago as the young lover of the infamous pirate queen, Captain Aurora. After Aurora’s disappearance, Lyda decided to stay on the island. Seeing the need and the potential for profit, Lyd decided to build a tavern, and it quickly became a popular gathering place for the locals, as well as for pirates who passed through.”
It wasn’t perfect. I had to edit, tweak, and create D&D-specific details for Lyd. I gave her a dwarven last name (Trueboots), determined her alignment, gifted her a few stats unique stats (+5 to Charisma checks), and figured out her personality traits. Still, by starting with the story created by the Craft AI Assistant, I turned what could easily take 30 minutes into a five-minute task.
From there, I popped into Discord, submitted a similar prompt to the Midjourney Bot that I gave to GPT-3 (“a dwarven woman who runs a tavern on a pirate haven”), and two minutes later, I had a portrait of Lyd to help my students bring her to life.
There’s a lot of controversy around these AI assistants. Artists and illustrators are rightly crying foul because the images the bots come up with have, at their origin, unique works of art that can be found on the Internet. They also work at a high enough quality now to replace the need for human illustrators and artists, and like the Luddites of yore, the skilled workers feel undervalued and dismissed.
Meanwhile, the word-generating AI bots will make teaching students how to write essays incredibly problematic. It’s too easy for an unscrupulous teenager to generate an original report on whatever topic their teacher asks for. These things can create titles (see the one that accompanies this post) or analyze the theme of The Scarlett Letter:
The theme of The Scarlet Letter is that of the consequences of sin and guilt. The novel examines how individuals respond to these consequences, and how societal norms dictate how we process guilt and the importance of justice. In the novel, Hester Prynne wears the scarlet A (for adultery) as a badge of her shame and is ostracized by her Puritan community for her actions. The novel examines themes of confession, repentance, and how the truth is often obscured by the judgment and prejudice of others. It also speaks to the power of secrecy and the importance of owning up to one’s mistakes in the face of public scrutiny. Ultimately, it reminds us that everyone is capable of sin, and that it is important to be forgiving and understanding instead of harsh and judgmental.
– Craft AI Assistant
There are real consequences to these technological advancements, ones that will not always be helpful to humanity.
I get that.
But I’m also a busy and stressed Dungeon Master, and if these things can make that role more manageable and fun, then I’m ready to roll.