Obsidian - Data Conversion with ChatGPT

Converting Pathfinder 2e (PF2e) stat blocks and items into Obsidian’s TTRPG stat block template format can be done quickly with ChatGPT—so long as the prompt includes the exact target template and the user iteratively corrects formatting mistakes. The payoff is substantial: instead of manually wrestling with YAML/front-matter syntax, a Dungeon Master can paste a monster (or even an NPC/item description) and get back a mostly usable Obsidian-ready block with the right structure for dice, actions, and encounter-style fields.

The workflow starts with the “official” conversion path: pull PF2e data from pf2e tools’ GitHub JSON files, then use a text converter to transform a stat block into JSON, and finally run it through the Homebrew converter to reach the Obsidian plugin’s expected format. That method works but involves multiple steps. The alternative—using ChatGPT—aims to collapse that effort into a conversational conversion.

ChatGPT’s early attempts often miss Obsidian’s strict syntax. Common failures include dropping the required front matter (the top “---” section), omitting the closing syntax at the end of the stat block, or outputting a different template variant than the one the user needs. The transcript shows a back-and-forth pattern: provide an example of the exact Obsidian template, paste a monster, request conversion into that specific format, then correct ChatGPT when it outputs the “wrong” structure. Over time, the model begins to follow the requested layout more reliably—especially when the user repeatedly anchors the conversation with the correct template and uses shorter inputs to reduce truncation.

A key practical limitation emerges: longer notes increase the chance of incomplete output, with ChatGPT sometimes stopping mid-block or failing to include the final closing lines. Even when the stat block is structurally correct, the content can still contain issues unrelated to formatting—such as duplicated monster data or mismatched HP values—so manual verification remains necessary.

Once the formatting is dialed in, the results look functional. A converted Web Loco stat block includes expected fields like perception, darkvision, languages, skills, speed, abilities, and action triggers, and it can be pasted into Obsidian notes with only light cleanup (like removing an extra “copy code” wrapper). The same approach works for other creatures (including an Owlbear and a Goblin Commando) when the prompt explicitly references the correct template and the conversation context is kept consistent.

The method extends beyond monsters. ChatGPT can convert magic items into the same Obsidian-compatible format—for example, turning a “bird feather token” into a structured item block with activation timing and rule text—though rule details may require review (such as activation duration or usage handling). Finally, the transcript demonstrates broader NPC generation: ChatGPT can draft a shopkeeper NPC with a list of items for sale and then format the inventory in a way that could be imported into Obsidian, potentially saving significant writing time.

Overall, ChatGPT is positioned as a time-saver for data conversion and content structuring in Obsidian, but it’s not plug-and-play. The conversion quality improves with template “training” through conversation, careful prompt specificity, and post-generation cleanup and validation.