This content originally appeared on DEV Community and was authored by cchsh
I have been building MDPR, a deterministic Markdown-to-editable-PPTX runtime, and mdpr-skill, an optional Codex companion for review hints.
The problem I wanted to avoid is common in LLM-generated slide decks: the model owns too much. It chooses coordinates, colors, emphasis, spacing, and object geometry directly. The first slide may look plausible, but coherence often drifts across the deck.
MDPR uses a different boundary:
- Markdown is parsed into presentation structure.
- Slide splitting and layout planning are deterministic.
- Theme colors, z-order, object bounds, and renderer output are owned by MDPR.
- The output is editable PPTX first, with HTML/PDF/PNG previews downstream.
mdpr-skill sits before that runtime. It can suggest compact semantic tags, icon-search keywords, Markdown cleanup notes, and visual QA concerns. It cannot own final coordinates, exact colors, z-order, shape geometry, exact icons, or renderer object IDs.
The short version:
The LLM can suggest; MDPR renders.
Install MDPR:
npm install -g @mdpresent/cli
mdpresent build deck.md --to pptx,html --out dist
Runtime:
https://github.com/ch040602/MdPr
Optional skill:
https://github.com/ch040602/mdpr-skill
npm:
https://www.npmjs.com/package/@mdpresent/cli
Preview gallery:
https://ch040602.github.io/MdPr/theme-preview/
I am looking for Markdown edge cases that usually break when converted into PowerPoint: dense tables, chart/table pairs, diagrams, mixed-language text, captions, and documents where editability matters after generation.
This content originally appeared on DEV Community and was authored by cchsh
cchsh | Sciencx (2026-06-24T10:09:53+00:00) Using LLMs for slide: MDPR + mdpr-skill. Retrieved from https://www.scien.cx/2026/06/24/using-llms-for-slide-mdpr-mdpr-skill/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.