Using Pandoc to write PDFs, Word documents, and slideshows in Markdown - Obsidian Community Talk

Pandoc turns an Obsidian-style Markdown file into polished outputs—Word documents, PDFs, and Reveal.js slide decks—by converting the content through a shared internal structure and then rendering it in the target format. The practical takeaway from this community talk is that once the command-line basics are understood (paths, input/output, and options), a repeatable workflow emerges: convert, add citations, apply custom styling, and automate the whole process inside Obsidian.

The session starts with command-line fundamentals that matter specifically for Pandoc usage. Commands follow a pattern: a program name plus arguments (like the input file) and optional flags (like output format or styling options). Just as important is “perspective” in the filesystem: absolute paths start from the computer’s root and work regardless of where the terminal is; relative paths depend on the current working directory. That distinction becomes the lever for shortening commands later.

With that foundation, Pandoc is introduced as a command-line converter that supports many document formats. Conversions follow a consistent structure: call `pandoc`, point it at an input file (including the extension so it knows the source type), specify an output filename and extension, and run. The talk demonstrates Markdown → Word and Markdown → PDF conversions using the same Markdown source. A key practical warning appears early: Pandoc overwrites output files if the target name already exists. Another concrete limitation shows up with embedded images: one embedding style carries over into Word and PDFs, while the other (the wiki-link style) does not.

Citations are handled through a three-part setup. First, citations are inserted into the Markdown using a site key (e.g., square brackets with an `@` prefix and the key). Second, Pandoc must be told where the bibliography data lives—typically by exporting a Zotero library to BibTeX or CSL JSON. Third, Pandoc needs a citation style definition (commonly APA) provided via a URL to a CSL style file. When those pieces are wired together, Pandoc generates both in-text citations and a reference list—though the reference heading must be added manually in the input.

For formatting beyond Pandoc’s defaults, the talk moves to “reference documents,” a Word-oriented template approach. The workflow is: export Pandoc’s default Word reference document, edit only the styles (not the content), then pass that reference document into the conversion command. The result is Word output with customized heading and paragraph styling.

Automation inside Obsidian comes next. The Snippets plugin lets users run shell commands from Markdown code blocks, so conversions can be triggered without leaving the editor. Chaining commands with `&` enables shorter relative-path commands by first switching into a common directory. Adding `start` opens the generated file immediately. Templater then layers on interactivity: it can pull the current file title/path, prompt for which citation style or reference document to use, and generate the final Pandoc command.

Finally, Reveal.js slideshows are produced from Markdown via Pandoc’s `reveal.js` output. The talk explains how YAML metadata becomes the title slide, how heading levels map to slide structure, and how options like theme selection, CSS injection, and reference handling can be combined. The closing Q&A highlights two workflow pain points: setting defaults to avoid repeating bibliography/style flags (via Pandoc’s data directory) and handling Obsidian block embeds (often better supported by an Obsidian Pandoc plugin, with command-line Pandoc as a fallback).