So I recently went to a seminar on how AI can help in the workplace. And it kind of inspired. It also really intrigued me to see what it could help me to do, if anything, in the RPG space.
I know this isn't new to many people, so I'll be brief.
I asked it to create an RPG adventure, for a specific RPG that was not DnD, about a ruined castle.
That was all I gave it.
The results were interesting. It was VERY fast. And I was impressed by the fact that the document it created opened with a set of hooks and a claim that this would appeal to different sorts of adventuring parties. I read them and I agreed, they would.
The AI also claimed that there were several possible endings, and that the adventure could be resolved in a number of different ways, by magic, negotiation or combat. Here I kind of agreed: it could be solved by magic, negotiation or combat - in the Inevitable Climactic Encounter which the characters could not succeed without having. But that, to be honest, is a typical DnD thing and really does depend on one's interpretation of 'different possible endings'. I didn't actually specify multiple endings, so the AI gets credit for partial success here, whatever you think about Inevitable Climactic Encounters. Another thing I did not specify was level specificity, but the document specified 3-5 characters of level 1-3 .
On the down side, the adventure was very generic. I mean SO generic. The title was 'Shadowheart Keep', which clearly is from Baldur's Gate III. The central figure (who may or may not be an antagonist) is Lord Ravencrest, a name that has at least been used before, in World Of Warcraft: Legion. So 'originality' is an issue, if you believe in that kind of thing (they are hardly 'original' names in the first place. The 'mook' NPCs were largely shadows and there were some haunted statues, dire ravens and other thematically appropriate stuff. But more importantly, you could play this game. Happily. It wouldn't be spectacular. There were only five locations, for a start, which is kind of limiting. And few surprises, if any, unless you were a beginner player.
At first, the document didn't include any stat blocks. When I asked for them, they came pretty quick, but were for 5e. I asked for something specific to the rules system I wanted to use and it said it didn't have access to that rule set. Which is cool. It's copyrighted, obviously. I resisted putting anything into the AI that wasn't my own and, in fact, didn't put anything that was my own in, either. Let's keep this relationship cordial for now... So I had to ask it to sniff around for something it could use and we settled on a good old 'OSR' system that fit the bill.
Then things got interesting.
I asked the AI to expand the adventure into a multipart 'mini campaign' and it did. In seconds, it extrapolated the premise of the original adventure, ret-conned it where it needed to, and created a four-part adventure. Complete with NPCs which it now statted. To be honest, part four was, in itself, a four-parter, and was pretty sketchy at that. Very brief. And my plan is to go back and get it to do that. You see my original preconception was that I would get a sketched outline or something, and then develop it up myself. But no. No, not at all.
Because I wanted to tackle that generic feel.
First, I asked it to expand the number of locations, based on what real castles were like, to include a well, and to create one subplot for every five rooms. It did that and now I know there was a conspiracy to undermine our Ravencrest character, long before the events that set this adventure in motion. There was already a plot of sorts to corrupt a spell he was trying to cast, which is why the adventure can exist. But now, there are a couple.
I then told the AI to rework some of the names. I got it to translate some of them into Old Norse, give me alternatives, explain how the names break down. Like this, I got rid of 'goblins' altogether.
The adventure is now looking very different from how it did originally; it has a different feel, a different implied setting, a bigger cast and stage. Now I've changed it all, I'll revisit the rest of the min campaign. I plan to visit those subplots and shore them up, see if I can't get some major consequences for the world written in there and even look back at the whole structure and ask myself if the Keep is even the best place to start.
And here's my big takeaway from all that:
It would have taken me tens of hour to do all this. As it was, it still did take hours. There's a lot of prompting etc involved in this. But. I can create a substantial campaign out of this in a fraction of the time it would have taken me on my own. I can keep my focus on my initial vision easily, because I don't have three hours of trying to assimilate a basic knowledge of Old Norse before I can give my characters cool names. I don't have to to and fro between and map and key, trying to strike the right balance between verisimilitude and fun. I don't have to spend hours draw a map, only to realise I need change it after I finish...
Anyway, if you want to see it, or a version of it, here's the link...
https://claude.ai/public/artifacts/a128284b-8b32-41bf-b39a-ade951754be2
No comments:
Post a Comment