Creating Generative AI CYOA Text-Based Campaigns with ASCII Maps and MidJourney AIart
Inspired by Obie’s post on Medium, My kids and I just played D&D with ChatGPT4 as the DM I’ve been exploring using ChatGPT for creating “dungeon crawler” roleplaying text-based campaigns. Also check out Steven’s post We asked ChatGPT to be the Dungeon Master in a game of D&D as well. Both are very good reads. But, what about the map where your character is exploring? What about imagery to visually show the characters within the campaign?
In this post I’m discussing how I expanded my personal use of ChatGPT with the following features.
1) a campaign from a Choose Your Own Adventure (CYOA) book,
2) an ASCII Map showing character movement, and
3) prompts for AIart using MidJourney.
All three of these can be prompts for ChatGPT. Using all three may add some complexity but also expand the experience with a generative AI CYOA text based RPG campaign. It would be interesting if there was an app that could do all three of these with ease :-)
As expected, you will read below that the experience was not perfect. For example, when attempting for ChatGPT to provide all three responses from one lengthy prompt was not very successful. Please share your feedback and ideas. I’d love to improve upon my process. Perhaps you have a suggestion for a 4th feature of a CYOA Campaign?
Feature #1 — Asking ChatGPT to Generate CYOA Campaigns
First, choose the Choose Your Own Adventure Story that you would like to go on an adventure. You will be surprised at how well ChatGPT will stay true to the story, but you may find times when the ChatGPT veers off course. Visit List of Choose Your Own Adventure books for a great list.
Here is an example prompt you can use.
Let’s go on an adventure. Please prepare an adventure the Cave of Time, Choose your Own Adventure book (written by Edward Packard) in 1979. The immersive story should provide my adventure choices in an A, B, C, format when decisions need to be made. I am a male teenager named Billy.
Feature #2 — Asking ChatGPT To Generate ASCII Dungeon Maps
Another aspect you can incorporate into your ChatGPT CYOA is an ASCII Map showing your character’s movement.
Ask ChatGPT to create you a dungeon map using ASCII text to tell you where you are and show your movement through your CYOA. Likely your CYOA is so immersive that a simple dungeon layout will not work. To me though it’s been fun to see ChatGPT try. For more information on this incorporation, check out my previous post, I survived a ChatGPT text-based adventure roleplaying game and lived to tell the tale.
Here are some of the responses I’ve received from ChatGPT.
Feature #3 — Asking ChatGPT to Generate AI Image Prompts for MidJourney
The final aspect you can incorporate into your ChatGPT CYOA is asking for visual prompts that you can copy/paste into /imagine prompts for MidJourney. I’d also suggest that you provide ChatGPT a little visual guidance. For example you can 1) briefly describe if your character’s age and gender, e.g., “please provide the young man,” 2) give your character a name, 3) ask ChatGPT to write the prompt in the third person, e.g., “written in the third person” which will help with the image generation and include the character, and 4) ask ChatGPT to generate your prompt following a “basic prompt anatomy.”
BASIC PROMPT ANATOMY
[STYLE], [CHARACTER], [SCENE].
STYLE = Paul Granger Color Illustration
CHARACTER = Billy is a male teenager wearing 1970s clothing and has brown hair
SCENE = Briefly describe your illustrative descriptive scene
Adding An Artist Name for Visual Style
For the style, consider describing the type of art and provide the name of an artist. For example, here are a list of some of the popular Choose Your Own Adventure illustrators.
- Paul Granger is one of the most iconic cover artists and illustrators of the CYOA series. He is known for his dynamic and action-packed illustrations, often featuring the protagonist in a dangerous or thrilling situation.
- Ted Enik had artwork typically showcased vivid and detailed scenes that capture the essence of the story, drawing readers into the adventure.
- Paul Abrams had art that featured bold colors and eye-catching designs. His illustrations often depict key scenes or pivotal moments from the adventure.
- Leslie Morrill had artwork that tended to be more subdued and atmospheric, often utilizing softer colors and evocative imagery.
The following is the prompt that I provided to ChatGPT asking for a Generative AI prompt.
Please provide me with a Generative AI prompt for the scene that is 1) written in the third person, 2) uses illustrative and descriptive language, and 3) follows the Basic Prompt Anatomy format.
BASIC PROMPT ANATOMY
[STYLE], [CHARACTER], [SCENE].
STYLE = Paul Granger Color Illustration
CHARACTER = Billy is a male teenager wearing 1970s clothing and has brown hair
SCENE = Briefly describe your illustrative descriptive scene
For example, here are some example prompts that I received and the AIart that was generated from MidJourney.
Generated Image #1
Color illustration in the visual style of Paul Granger, Billy is a male teenager wearing 1970s clothing and has brown hair, Billy is falling through a whirling, swirling vortex of color and light and crash-lands on a solid surface in an unknown location.
Generated Image #2
Vividly painted sci-fi mural in the style of Paul Adams, Billy male teenager wearing 1970s clothing with brown hair, sees a strange insect-like creatures buzzing past him, their wings are a blur of color as he step onto a new alien world with a deep royal purple sky.
Asking MidJourney to /describe Paul Granger Illustrations
The generated AI images above are nice, but not quite the look I was going for. What I decided to do was upload some images from the Cave of Time Choose Your Own Adventure Book written by Edward Packard and illustrated by Paul Granger in 1978. Using the /describe prompt, MidJourney will provide suggested text prompts (included below each image example).
/Describe Paul Granger 1978 Illustration 1
A sketch shows a person at a point looking toward the mountains, in the style of inuit art, stereoscopic photography, romantic riverscapes, expressive comic panels, 1918–1939 (interwar), panorama, coastal scenery.
Using this description as a prompt, I received.
/Describe Paul Granger 1978 Illustration 2
Dragon with mouth in water from the old library book, in the style of frank thorne, distorted and elongated forms, george herriman, lovecraftian, einar hákonarson, charming illustrations, elongated and dramatic — ar 81:133
Using this description as a prompt, I received.
/Describe Paul Granger 1978 Illustration 3
a black and white illustration is made of an animal with its mouth open, in the style of adventure pulp, 1970–present, cryptid academia, manticore, children’s book illustrations, bold shadows, art of the upper paleolithic — ar 81:85
Using this description as a prompt, I received.
/Describe Paul Granger 1978 Illustration 4
hobbit, chapter 1, openend illustration, in the style of black mountain college, chinapunk, rough hewn surfaces, realistic attention to detail, blink-and-you-miss-it detail, greeble, text-heavy — ar 95:72
Using this description as a prompt, I received.
/Describe Paul Granger 1978 Illustration 5
a black and white old engraving of a person riding a horse in front of a castle, in the style of whimsical children’s book illustrator, new leipzig school, comic strip, harry watrous, ultrafine detail, jump cuts, suburban ennui capturer — ar 182:81
Using this description as a prompt, I received.
/Describe Paul Granger 1978 Illustration 6
fairy kingdom: story of fairy horses and princesses / a fairy tale, in the style of black-and-white block prints, pont-aven school, comic satire, realist detail, jump cuts, the blue rider
Using this description as a prompt, I received.
Update Prompts with MidJourney /describe Medium & Style
The generated AI images above are closer to what I was looking for, but not for the visual descriptions I received from ChatGPT. Now I wanted to apply that previous description to these /describe prompts. Please note that I did remove erroneous information from the MidJourney prompts that I thought did not ADD to the AIart. I selected this one because I thought it was the closest to the illustrations within the book.
a black and white old illustration of a person riding a horse in front of a castle, in the style of whimsical children’s book illustrator, new leipzig school, comic strip, harry watrous, ultrafine detail, jump cuts, suburban ennui capturer
MidJourney /describe Style + ChatGPT Scene Example #1
A black and white old illustration of Billy who is a male teenager wearing 1970s clothing and has brown hair, Billy is falling through a whirling, swirling vortex of color and light and crash-lands on a solid surface in an unknown location in the style of whimsical children’s book illustrator, new leipzig school, comic strip, ultrafine detail, jump cuts.
MidJourney /describe Style + ChatGPT Scene Example #2
A black and white old illustration of Billy male teenager wearing 1970s clothing with brown hair, sees a strange insect-like creatures buzzing past him, their wings are a blur of color as he step onto a new alien world with a deep royal purple sky, in the style of whimsical children’s book illustrator, new leipzig school, comic strip, ultrafine detail, jump cuts
Advanced Prompt Guidance
Check out Tristan Wolff’s Midjourney Cheat Sheet -V5 for more detailed information and consider using the following to create your MidJourney Prompts. As I continue to use the process above I’ll be using Tristan’s resource for ideas.
Where Should I Go Next?
As expected, the experience was not perfect. For example, when attempting for ChatGPT to provide all three responses from one lengthy prompt was not very successful. Please share your feedback and ideas. I’d love to improve upon my process. Perhaps you have a suggestion for a 4th feature of a CYOA Campaign?
