I made a rigorous AI Dungeon Master/Creator for open-ended D&D 5e play — looking for testers
Hey everyone,
I’ve been building a custom AI Dungeon Master for D&D 5e, and I’m looking for people willing to test it and give feedback.
The model is designed to do two main things:
First, it can create structured D&D adventures from a character sheet, backstory, campaign notes, loose idea, recap, or uploaded files. Instead of just writing a plot summary, it builds something closer to a DM-facing adventure framework: a premise, opening situation, major scenes, NPCs, clues, branching paths, combat or social encounters, consequences, and a climax that can shift based on player choices.
Second, it can run the adventure as a live AI Dungeon Master. It is meant to handle open-ended player actions instead of forcing fixed menu choices. You can investigate, negotiate, sneak, fight, use spells creatively, ask NPCs questions, pursue alternate routes, or derail the expected path, and it should try to adapt while keeping the session coherent.
The model is built to avoid some common AI-DM problems, like:
giving vague atmospheric descriptions with nothing useful to interact with
asking “what do you do?” before giving enough context
turning every scene into passive narration
forgetting NPC goals, clues, consequences, or clocks
making combat too abstract
ignoring D&D 5e mechanics
forcing the player down one intended path
making every NPC sound the same
expanding a one-shot into a sprawling campaign
It can also help outside of live play. You can use it to:
turn campaign notes into a playable session
create a one-shot for a specific character
continue from a previous session recap
build a backstory-focused side quest
design a mystery with clues and multiple routes
create tactical combat encounters
diagnose pacing problems in an adventure
revise an existing adventure so it has better structure and player agency
The overall goal is for it to feel more like a prepared but flexible DM: structured enough that the game does not drift, but open enough that player choices actually matter.
I’m especially looking for feedback on whether the adventures feel playable, whether it gives enough context before asking for decisions, whether combat and investigation feel usable, and whether it handles open-ended choices better than a normal chatbot.
And yes I had it write this post for me.