Obsidian - Create Random Notes with ChatGPT Templates

Random tables are a staple of tabletop RPGs, but maintaining them is tedious—especially when new books and options keep arriving. This workflow uses Obsidian’s Text Generator plugin to generate fully formed RPG content on demand by sending carefully written templates to ChatGPT, turning “I need a shop/inventory/encounter/rule-of-thumb” into a quick Ctrl U or Space Space trigger.

The core setup starts with installing the Text Generator plugin from Obsidian’s Community Plugins, then disabling Restricted Mode so plugins can be installed. After installation, users must add an OpenAI API key from platform.openai.com. The guide emphasizes cost control: OpenAI billing settings include usage limits, and failing to set a hard monthly cap can lead to surprise charges. Once the key is pasted into Obsidian, the plugin is configured to use a model—GPT 3.5 turbo is recommended for lower cost, while GPT4 is described as more expensive and access-restricted.

Configuration then focuses on output quality and template behavior. Max tokens is increased (the presenter lands around 650 to avoid errors seen at higher values), and temperature is raised to 0.8 from 0.7 to make results more varied. The plugin’s prompt/template folders are pointed at a local directory (text generator slash prompts), and options like including front matter and headings are enabled so generated content fits neatly into Obsidian notes. A key quality-of-life feature is enabling token/cost estimation for templates, which surfaces an “estimate tokens for a template” option in Obsidian’s command palette.

Once the plugin is working, the real payoff comes from templates. Templates live in the Text Generator templates package structure and include front matter fields such as prompt ID, name, description, author, tags, and version. The template “name” is what appears when triggering templates via Ctrl U, so emojis can be used to make selection easier. The guide’s examples show how simple prompt instructions can produce structured RPG assets: an inn template generates an innkeeper profile, signature drinks in a Markdown table (name, cost, description), meals with prices, patrons, and a rumor hook. An encounter template generates a Dungeons and Dragons 5th edition encounter for a specified party size and level, then outputs an encounter block plus descriptive text.

To keep encounters compatible with downstream tools, the template instructions include strict formatting rules—such as requiring monster names to match exactly what’s in the monster manual and avoiding pluralization that can break initiative tracking. The workflow also supports context injection: highlighting text in a note and triggering the template lets the generated content match the current scene (e.g., “land of lollipops and candy canes” becomes a themed encounter).

Beyond inns and encounters, the same method is used to generate job boards, adventures, dungeons, shops, apothecaries, armories, and even Mythic GM emulator-style prompts like a Magic 8 Ball. The guide also notes system-agnostic potential: as long as ChatGPT has enough information about a ruleset, templates can be adapted for other RPGs. For localization concerns, the guide suggests keeping offline backups, and it highlights a practical expansion path by generating items sourced from Cobalt Press through the OGL.

Finally, templates are shared via Patreon, and the cost is framed as manageable—testing reportedly totals around 12 cents—while the plugin’s token estimation helps users stay aware of spend. The overall message is that templated AI generation can replace years of manual random-table upkeep, letting GMs spend more time on play and less time building content lists.