Markdown to Beamer: LaTeX slides from the outline you already have.

Beamer is the LaTeX class behind most conference and lecture slides in the sciences. Its output is excellent — real maths, consistent typography, a PDF that projects the same everywhere — and its input is tedious: every slide is a frame environment, every list an itemize, and a slide with code fails until you remember [fragile]. Yet the content of a talk is already an outline. Writing that outline in Markdown and letting a converter produce the Beamer source keeps the output and drops the boilerplate.

The two-minute version

1. Write the talk as Markdown: a YAML header for the title page, # headings for sections, ## headings for slides.
2. Paste it into the Markdown to LaTeX converter and switch to Beamer.
3. Read the frame preview on the right; each card is a slide, and clicking one jumps to its line in the source.
4. Choose a theme and aspect ratio from the Slides menu, then download the .tex or click Open in Overleaf.

How headings become frames

Everything hinges on one setting, the slide level: the heading level that starts a new frame. Headings above it become sections, headings below it stay inside a frame. With the level on Auto the converter applies Pandoc’s rule: the slide level is the shallowest heading that is followed directly by content rather than by another heading.

In practice that gives four rules worth remembering:

• Above the slide level — usually # — a heading becomes \section (or \subsection). It appears in the outline and in navigation bars, but makes no slide of its own.
• At the slide level a heading starts a frame and becomes its title.
• Below the slide level a heading opens a Beamer block inside the current frame, which is how you get a titled box on a slide.
• Content before any slide heading still gets a frame, without a title, so nothing is dropped.

Auto suits almost every outline. Fix the level when you want the structure read differently — at level 1, each # section becomes one slide and its ## headings turn into blocks; at level 3, every ### gets a slide of its own.

The title page comes from YAML

A YAML block at the very top of the file fills Beamer’s title fields, and the converter adds a \titlepage frame first:

---
title: Quarterly review
subtitle: Q3 2026
author: Ada Lovelace
date: October 2026
---

# Results

## Revenue

- Up 12% on last quarter
- Margin $m = 0.31$

## The code

```python
print("fragile frame")
```

---

Questions?

title, subtitle, author (a single name or a list), date and institute are read. Without a YAML title, a single # heading on the first line is taken as the title instead — handy for an outline that starts with its name. The Outline frame option adds a table of contents after the title page, and Section pages adds a divider slide at every section.

Breaking a slide without a heading

A horizontal rule — --- on its own line — starts a new frame with no title, which is what you want for a full-slide quotation, an image, or a closing “Questions?”. Leave a blank line above it. Markdown reads --- directly under a line of text as an underline that turns that line into a heading, so the blank line is what makes it a break. If you use rules as decoration instead, switch off --- starts a new frame.

Code on a slide needs [fragile]

This is the error that catches everyone writing Beamer by hand. Beamer collects a frame’s body as a macro argument before it typesets it, and verbatim environments such as lstlisting cannot live inside an argument: they need to change how TeX reads characters, which is too late by then. The fix is the [fragile] option, which makes Beamer write the frame to a temporary file and read it back.

The converter adds [fragile] to every frame that contains a fenced code block and to none of the others, which is also what Pandoc does. Tag the fence with its language — ```python — and the listing is highlighted; the listings package knows Python, C, C++, Java, Bash, SQL, R, MATLAB and a few dozen more.

Math, tables and images on a slide

Maths is copied into the frame exactly as typed, so $…$ and $$…$$ behave as they would in a paper. Tables become booktabs tabular environments wrapped in adjustbox{max width=\linewidth}, so a wide table shrinks to fit the slide instead of running off it. Images are centred and capped at about two thirds of the slide height, and each is guarded by \IfFileExists: if the file is missing the slide shows a framed note and the deck still compiles. Use PNG, JPG or PDF — LaTeX cannot include SVG or WebP directly.

Choosing a theme

Beamer ships with dozens of themes; the converter offers the seven people actually use. The same frame in six of them:

metropolis is the default here because it is flat, high-contrast and readable from the back of a room. It uses the Fira fonts, so the converter picks XeLaTeX for it; with pdfLaTeX it still compiles in a fallback font. Madrid is the classic look, with a coloured title bar and a footer carrying author, title and page number. Berlin adds a navigation bar of sections and dots at the top. For every theme except metropolis the converter removes Beamer’s navigation symbols, which almost nobody clicks. Pick 16:9 for modern projectors and 4:3 for older rooms.

Revealing a list one item at a time

Turn on Reveal lists item by item and each list is written as \begin{itemize}[<+->]. Beamer then shows one more item per click. Each step is a separate page in the PDF, so a five-item list becomes five pages — expected, and the reason to use it for the slides where pacing matters rather than everywhere.

When a frame is too long

A slide holds far less than a page. The converter estimates the height of each frame and adds a note when one looks likely to overflow, naming the frame. Splitting it with another ## heading or a --- is almost always better than shrinking the text; if the content really is one continuous list, add [allowframebreaks] to that frame in the .tex and Beamer will continue it on a second slide.

The same deck with Pandoc

Because the converter follows Pandoc’s rules, the Markdown you write here also works on the command line, which is the better choice for a deck rebuilt in CI or kept with its data:

pandoc slides.md -t beamer -o slides.pdf \
  --slide-level=2 \
  -V theme:metropolis -V aspectratio=169 \
  --pdf-engine=xelatex

Add -i for incremental lists. Pandoc also understands a few extensions the browser converter does not yet handle: ::: columns for two-column slides, ::: notes for speaker notes, and . . . for a pause. If you need those, start here to get the structure right, then move the Markdown to Pandoc.

Compiling: pdfLaTeX, XeLaTeX or Overleaf

The status bar under the preview names the engine the .tex expects. Most decks compile with pdfLaTeX; metropolis prefers XeLaTeX; and any Chinese, Japanese or Korean text switches the class to ctexbeamer with the Fandol fonts, which also needs XeLaTeX. If you have no TeX installation, Open in Overleaf creates a project with the right compiler already selected, and images referenced by URL are sent along with it.

Writing a paper rather than a talk? The same converter has an Article mode, and Markdown vs LaTeX explains when drafting in Markdown pays off.

Frequently asked questions