Google NotebookLM for RPGs - Making my own Virtual Campaign and DM assistant

Black Dougal

Footpad
Hey Enworlders,

Long time lurker and occasional commenter here. I believe this might be my first post.

I've been seeing a lot of "what do you use ChatGPT for" and "how to use AI for your RPG games and campaigns". I thought I'd share how I use Google NotebookLM to help me as a DM. How people can use it of course is up to them. I imagine like ChatGPT and the like there a gazillion ways you could use it. I use it as a uber massive DM Screen/Virtual Assistant.

So for those who are not familiar, Google NotebookLM is presented as a research tool/assistant. In good news, the FREE VERSION should suffice.

Details:
Google NotebookLM is an AI-powered research and note-taking tool developed by Google Labs. It leverages advanced language models, specifically Google Gemini, to assist users in interacting with their documents. By uploading various sources—such as Google Docs, PDFs, web URLs, and YouTube videos—users can engage with their content more deeply through features like summarization, question answering, and content generation, all grounded in their personal materials.

Key Features:
  • Source Integration: Users can upload up to 50 sources per notebook, including documents, slides, PDFs, web URLs, copied text, and YouTube URLs. This allows NotebookLM to analyze and connect information from diverse formats.
  • Summarization and Q&A: NotebookLM can generate summaries of uploaded content and answer user queries, providing responses with inline citations that link back to the original sources.
  • Content Generation: The tool can create various outputs based on user sources, including study guides, briefing documents, timelines, FAQs, and audio overviews.
  • Audio Overviews: Introduced in September 2024, this feature generates podcast-style discussions between AI hosts about the user's uploaded material, offering an engaging way to understand complex information.

So, unlike ChatGPT and many other LLMs it uses your specific data. Considering many campaigns and games are homebrewed or tailored to our play tastes and events in games themselves it is immensely useful. It's a data model specific to you and yours.

I use it to help with Lore, rules, campaign and session notes and any other DM related tasks while running games. Essential my own virtual assistant referencing my pre determined library of books.

Example:

-I have a Core Rules and Current Adventure Notebook. (all my DM, Player, rules supplements, homebrew rules, and Monster pdfs and currently The Lost Caverns of Tsojcanth)
-I have a weekly session to session notes Notebook. (I track various campaign elements then use it to create notes I feed into my Campaign world Notebook as well.
-I also have a campaign world Notebook. (46 sources of Forgotten Realms pdfs from all editions, as well as our homebrewing notes and campaign diaries. We've been playing in our version of the FR since 87 on and off. Just to give you an idea of scope of how specific/homebrewed my FR is to canon. That's a lot to try to remember.)
- Player/PC Notebook (loaded with pdf character sheets, notes, PC diaries, etc.)

This allows me to reference rules, campaign info, world info, etc etc for my games specifically in seconds.

"What is the capital of Tethyr?"
"What are the overland travel rules for mounts"
"Create a random table of 20 forest based monsters who are also aberrations between CR 1/8 and 4, also, only evil alignments"
"How did [insert player] acquire the Vorpal Sword?"
"Outline the Society of the Horn include, history, motivation, leaders, etc."
"Create a list of 20 random rumors one might hear in a tavern in Westgate"
"create player facing lore for the Time of Troubles"

The dice have determined a random planar gate exists in the hex, another roll determines it a gateway to the 422nd layer of the Abyss.

"Give me an detailed outline of the 422nd layer of the Abyss"
"Create a list of 12 random monsters related to the Abyss as well as the 422nd layer of the Abyss specifically."

It also allows me to create "podcasts" of varying nature for my players. Recap podcats, lore podcasts, etc.

I'd love to hear more about how some of you use Google NotebookLM. Also, feel free to ask questions if you have any.
 

log in or register to remove this ad




How often does it get things wrong?

Not much. It's not flawless, of course. I find at times I have to ask it to redo an answer due to format of answers but not much on the side of making mistakes or hallucinating.

So for example I may ask it to create a list of al goblins for a wandering monster list, it might screw the formatting a bit now and again or present 10 types of goblins instead of the full 15, for example.
 

Not much. It's not flawless, of course. I find at times I have to ask it to redo an answer due to format of answers but not much on the side of making mistakes or hallucinating.

So for example I may ask it to create a list of al goblins for a wandering monster list, it might screw the formatting a bit now and again or present 10 types of goblins instead of the full 15, for example.
I find more that it has problems because I haven't provided the context in my own sources, but I asked it a question assuming that contextualization was there. Or I was ambiguous on some relationship between sources, and it used the wrong one.
 

Yeah, it's not perfect but it helps me run my open ended sandboxes quite nicely. I've also made my own GPTs for random hex generation based on criteria I want and only dealing with my homebrew lore etc. Between the two it works nicely with the sandbox mentality and helping me improv quickly as needed.
 

So, in another thread, I asked if there was a tool to compile soccer stats for a team I'm coaching.

I tried it and it worked perfectly. I uploaded an audio play-by-play and then asked it to create a spreadsheet with all the player's stats - I listed the stats I wanted (goals, successful passes, interceptions, assists etc...)

It created the spreadsheet and I saved it as a note. I can even copy/paste it into a word doc in order to print it out.

I find more that it has problems because I haven't provided the context in my own sources, but I asked it a question assuming that contextualization was there. Or I was ambiguous on some relationship between sources, and it used the wrong one.
So I noticed this too. I started trying it out (after I was done with my soccer experiment) for my game by inputting all the random stuff that people have been posting (synopese, magic items etc..)It doesn't really understand the chronological order of things so I would have to be a lot more careful on labelling when things happen. I'm not sure how to do that though. Maybe by starting by dating each source? Maybe then it will figure out the order of events better.
 


Remove ads

Top