Convert a White Paper from Markdown to a Serif Academic PDF

A benchmark table renders for technical readers

The credibility centrepiece of many white papers — a benchmark table in GFM pipe syntax becomes a bordered table with a shaded header row.

Markdown source:
## 4. Benchmarks

| System   | p99 (ms) | Throughput |
|----------|----------|------------|
| Baseline | 142      | 8k rps     |
| Proposed | 61       | 19k rps    |

PDF output:
4. Benchmarks  (serif 1.5em, underlined)
[ bordered table, gray header, serif body ]

Complexity notation prints literally

Technical white papers often cite complexity; dollar-delimited math is not typeset, so write it as plain text.

Markdown source:
The index lookup is $O(\log n)$ amortised.

PDF output (literal):
The index lookup is $O(\log n)$ amortised.

Fix: write 'O(log n) amortised' as plain text, or include the
expression in a figure.

A config block stays readable but uncoloured

Useful for reference architectures: a config or code block keeps its monospace box, but reviewers won't get syntax colours.

Markdown source:
```yaml
replicas: 3
strategy: rolling
```

PDF output:
[ gray #f6f8fa box, monospace 0.85em, no colour ]
replicas: 3
strategy: rolling

Note: enable 'Background graphics' or the box prints white.

References render as an indented list

An ordered references list comes through indented, giving the white paper academic credibility for B2B audiences.

Markdown source:
## References
1. ACME Labs. (2025). *Throughput study*.
2. Doe, R. (2024). *Index design*. Internal report.

PDF output:
References  (serif 1.5em, underlined)
1. ACME Labs. (2025). Throughput study.
2. Doe, R. (2024). Index design. Internal report.

Strip the version/metadata header

White-paper sources often start with a YAML block of title/version/author; remove it or the PDF opens with a broken heading.

Markdown source:
---
title: Scaling the Index
version: 1.2
classification: confidential
---
# Scaling the Index

PDF output (broken):
[ horizontal rule ]
title: Scaling the Index version: 1.2 ...  <- garbled H2
Scaling the Index

Fix: delete the --- block (manage metadata separately with
md-frontmatter-builder).

$O(n)$, $\lambda$, and $$…$$ print as literal text — there is no math engine. Write expressions as plain text or include them in a figure. md-math-normalizer only standardises delimiters for a downstream renderer.

[^1] markers and definitions print verbatim. Convert to inline citations, or run md-footnote-linker first.

A leading --- YAML block becomes a horizontal rule plus a garbled heading — and may print a 'confidential' classification you meant to keep out of the body. Strip it before exporting, or manage it with md-frontmatter-builder.

Fences get a gray box and a language-* class but no colourising library runs. Config stays readable as monospace; for highlighted code in a web context, use md-to-html.

This produces a print-styled PDF, not an editable file. For a Word document marketing can restyle, convert with md-to-docx instead.

A document over the plan's character cap (500,000 on Free) is refused with the count and limit named — often from embedded base64 diagrams. Use external image URLs or upgrade for longer papers.

A relative image src can't resolve in the preview frame. Use an absolute https:// URL, or rewrite paths with md-image-path-rewriter.

Browsers omit print backgrounds by default. Enable 'Background graphics' / 'Print backgrounds' so benchmark-table headers and code boxes appear in the client-facing PDF.

The tool takes one file per run. Merge sections with md-merger into a single document first, then convert.

The tool styles your paper into a print frame; the .pdf is produced when you choose 'Save as PDF' in the browser dialog. There is no server-side renderer — which is why proprietary research stays local.