T4zzM4nn's AI Dungeon - S01E40 - Story Cards AI Generation Prompts

AI Dungeon Story Card Generation

The episode explores new AI-generation features for AI Dungeon Story Cards, with a brief mention of upcoming Voyage content.

Prompts and templates

Story Card generation offers four prompt templates: Basic Character, Basic List, Basic Prose, and Custom Command.

The Basic Character template requires the card title as a prompt token and asks the AI to generate labeled fields such as name, role, species, age, gender, personality, and appearance. It has a 750-character limit.

Users can edit prompts and save or update custom versions for reuse. The host found no built-in location template, but demonstrated adapting a prompt to generate a concise card for a small city, including points of interest and demographics.

To generate a card, set the desired command, return to the card details, and select Generate with AI. Generating again appears to replace the existing content rather than append to it.

Formatting and token tradeoffs

The host compared plain text, curly-brace grouping, and square-bracket grouping. He speculated that grouping might help the AI associate details with one character, but did not establish that it improves interpretation.

Labeled fields are easier for people to scan, while compact prose can use fewer tokens. The video recommends balancing readability against token efficiency, especially when conserving tokens. In one example, rewriting a character card as compact prose reduced its length while retaining key details.

Other interface change

A new Demographics section in settings lets users enter details such as age and gender. The host was unsure how this information would be used.

Reviews, guides, and announcements

This is a hands-on walkthrough and experiment rather than a formal review. It demonstrates creating character and location cards and customizing generation prompts.

Voyage is mentioned as an upcoming feature that the host planned to cover in later videos; it is not demonstrated here.

Main speaker/source: T4zzM4nn, the video’s host and demonstrator.