A brain cell-popping contribution from Captain’s Log super fan Mike Yacullo.

I’ve been a Star Trek fan as long as I can remember – I grew up watching TOS reruns with my brothers, my parents, and my aunts in the 70s. (Shout out to WPIX in New York!) I’ve been playing RPGs since the D&D Basic Set came out in 1977.

Star Trek Adventures had been on my radar for a while, but I’d never had the chance to play it. When I heard about Captain’s Log, I was very excited to give it a spin. I’d done a ton of solo board gaming, and some solo RPGing (mostly with Ironsworn and Starforged) so this looked like it was right up my alley.

As I was establishing the scenario for my first game of Captain’s Log, I had a lot of questions to answer. Who was the rest of my crew? What about the crew we were trying to rescue?

This was around the same time I was getting really interested in GPT-4, so I thought it was a good opportunity to see what it could do for my game. Note that this was done using a paid GPT-4 subscription, so using a different engine may give you different results.

I started simply, just trying to establish some “sketches” of crew members – names, descriptions, and traits that would give my character something to work with. After some trial and error, I had a prompt that would reliably generate an NPC crew:

The results were quite good, although I did make some minor changes – for example, I would not be calling Captain Isabelle Grant by her nickname “Izzy” that GPT-4 suggested! This is typical of AI-generated content – it’s often very good, but rarely perfect. When working with AI, don’t feel constrained by the results the engine gives you. I usually copy the results into another document and make the changes I want.

I’m not a writer, and this is just for my own enjoyment, so I write up my Captain’s Log games in a screenplay-ish format. These rough characters gave me a much better mental image to work with and really helped me get moving.

As my game went along, I found myself wanting more detail for the crew members. Although I like the mechanic of using the ship’s System and Department rating to resolve actions, it didn’t feel right to me in the context of a crew member on a landing party attempting an action. After using the Yes/No matrix a few times, I decided I wanted stats for my crew NPCs. Back to GPT-4!

This one took some trial and error. On my first try, I started by saying:

and then pasted in the descriptions generated in the previous chat. The results were decent but showed relative rankings instead of numerical stats.

Again, this is fairly typical of working with AI, getting close and having to nudge in the right direction until you get things the way you want. Once I had it generating the numeric stats, I had to explicitly tell it to put the Attributes and Disciplines in the order they appear on the character sheet. Eventually, I ended up with a prompt that worked the way I wanted:

I was very happy with the results from this prompt. Now I had fully fleshed-out characters I could make action rolls for. As usual, the AI does not always follow instructions 100%, so keep your eyes on the results and either regenerate or make changes manually.

Once I have a prompt I want to reuse, I add it to a document so I can easily find it again without scrolling through all my trial-and-error chats.

At this point I was ready to combine all the previous work into a single mega-prompt to fully generate a crew whenever I need one:

Lastly, I wanted to generate images for my crew to complete the mental image. Now that DALL-E is integrated into GPT-4, it was simple to ask

and the results were mostly quite good.

For the characters I wasn’t satisfied with, I prompted individually until I had the results I wanted. It’s really nice to be able to just speak to GPT/DALL-E rather than do the much more technical prompting a tool like Midjourney requires. You simply ask for features to be added, changed, or removed. Here are some before-and-afters, with the prompts I used to change one to the other.

There’s plenty of room for experimentation here. Depending on the flavor of your campaign, you may want to see illustrations in different styles such as “concept art”, “anime”, “star trek the animated series”, and so on. Here’s Dr. T’Vala in a number of different styles.

A few problems I haven’t been able to crack:

  • It doesn’t seem to understand a lot of the Trek alien races. My main character is Bolian, and while I got a lot of blue-skinned humans with various forehead bumps, it did not know to give the Bolian the distinctive center line even though I tried to prompt many different ways to get it. Trill spots, Andorian antennae, and other races’ distinctive features are hit-or-miss as well.
  • I wasn’t able to get it to make the uniforms consistent. Even when specifically asking for TNG or TOS uniforms, each character seemed to have their own unique uniform style.

For my purposes, this was fine – I just wanted to have a mental image of each character. If I was going to publish my adventures somewhere, I would spend more time refining the image prompts and probably doing some Photoshop to get everyone into the same uniforms.

Using GPT-4 in this way has really enhanced my gaming experience – letting me quickly generate characters for my main character to interact with and giving them enough detail that I have a good idea who they are. When I run into another ship, boom, I instantly know the name of the captain or any other crew member my crew needs to interact with.

I would love to hear about improvements to these prompts, or other successful approaches people are taking.

Thank you for reading, LLAP!


TM & © 2026 CBS Studios Inc. © 2026 Paramount Pictures Corp. STAR TREK and related marks and logos are trademarks of CBS Studios Inc. All Rights Reserved.

14 responses to “Charting the Stars with ChatGPT-4: Crafting Star Trek Adventures”

  1. Prompts worked well, good work, it seems to be poorly responding on the majority of the crew needs to be human portion. I would set a limit of 2 characters that are not human as part of the propmt. Only thing to note is that utilizing the DALL-E interface requires payment of $20 a month and is not part of the free service.

  2. I regularly use GPT-4 for my writing. Not to actually write, but to help me plot out the details in my novels and organize my thoughts. Also, it’s a great way to bounce ideas off of something that doesn’t get annoyed (i.e., my wife). Since I received Captain’s Log for my birthday, I’ve been using it to bounce ideas off for getting everything set up for my “game.”

    One thing I love about GPT-4 is how easy it is to come up with side-characters and come up with plot formatting. I have used a similar set of prompts as above to flesh out an attack fighter squadron in my Top Gun-style Captain’s Log. It’s even given me a couple secondary plot points that I never thought of. In the end, it’s a good tool that can be great for Captain’s Log.

  3. I have been using multiple program to make authentic TOS costumes, male and female.

    1. Oh, I’d love to hear about that!

  4. Did you have to first feed it the STA Captain’s Log book or did it generate all this with the default context it was trained on?

    1. This was all from the default context. I should have mentioned in the article that I did start by asking it to confirm it was familiar with Captain’s Log.

      1. Hello Mike,
        I’m Edward Benson. I read your article, and would love to learn more about this process. I am just now experimenting with ChatGPT and really need to get to a better level of operating with it. Which one did you use? How do you first ask if it’s familiar with something, before generating an image? I thought it didn’t communicate and only generated images? Thank you for your time, and happy holidays!

      2. Hi Edward! I’m using GPT-4, which costs $20/month. You can find it here: https://chat.openai.com/

        You really do just ask it “are you familiar with xyz”. Usually it is, and you’re off to the races. If not, you can provide it with as much detail as you need to to get it going.

        The images were also generated with GPT-4, which is now integrated with the DALL-E image generation engine. This is great because you can do it all within one conversation, using the text output as a prompt for the images.

        Hope this is helpful, good luck!

    2. This is from the default context. I should have mentioned that I begun by asking it to confirm it was familiar with Captain’s Log and Star Trek Adventures.

  5. I’ve just started using ChatGPT 3.5 and Midjourney for something similiar. I’m going to try adjusting my prompt for ChatGPT based on what you have done to get the formatting a little better.
    Only tried a couple of the characters in Midjourney, but getting the uniforms right does seem to be very challenging.

  6. Hello All. I admit to being a bit of a N00b when it comes to the various ChatGPT applications. I have played with free versions, but I am at a loss as to understand where (where, or which, or how) you have a conversation with it to determine if it was familiar with Star Trek Adventures. I would like to ask which platform all these great images where generated with, and how much the “cost” is please.

  7. Edward, see my response under your comment above.

Leave a Reply to Mike YaculloCancel reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Discover more from Continuing Mission

Subscribe now to keep reading and get access to the full archive.

Continue reading