DocsFastFormat support — the honest matrix
Every converter has limits. Most hide them. This page documents ours per element and per direction, generated from the same capability table the tools report at conversion time.
Last verified: 2026-10-25 · engine v0.5.0
Each verdict below is generated from the same data the fidelity report uses at conversion time — Native means the element maps to a real equivalent in the output file, Approximated means you still get something useful but not the full-fidelity form, and Dropped means it is omitted (and called out in your report). Nothing here is a silent transformation.
Markdown → PDF
Exported with the Markdown to PDF tool. 13 element rules; highlights: tables whose header row repeats across page breaks, clickable links, A4 with page numbers.
| Element | In the PDF |
|---|---|
| Headings H1–H6 | Native — Sized and bolded on the page (PDF outline/bookmarks not generated yet) |
| Bold / italic / strikethrough | — |
| GFM tables | Native — Real table with a shaded header row that repeats across page breaks |
| Links | Native — Clickable, coloured in the PDF |
| Bulleted lists, nested | Native — True numbered and bulleted lists, nested |
| Ordered (numbered) lists | Native — True numbered and bulleted lists, nested |
| Task lists (checkboxes) | Native — Printed checklist glyphs ☑/☐ — the correct form for a static PDF |
| Blockquotes | Native — Indented, italic block styling — the typeset form of a print blockquote |
| Fenced code blocks | Native — Monospace with a shaded background and preserved spacing |
| Images | Approximated — data: PNG/JPEG images are embedded; remote/relative files are never fetched (privacy) and stay flagged in the report |
| Footnotes | Approximated — Collected into a Notes section at the end of the PDF (not paged footnotes) |
| Math (LaTeX) | Approximated — Typeset with real sub/superscripts, √ and ∑/∫ glyphs (engine v0.4.0); stacked fractions and OMML remain Word-only |
| Raw HTML blocks | Dropped — Removed for safety |
| Front-matter (YAML) | Native — title/author/subject/keywords map to the PDF document info; other keys are listed in the report |
| Mermaid code blocks | Approximated — Rendered as code blocks; diagram rendering is planned |
Markdown → Word (DOCX)
Exported with the Markdown to Word tool. 15 element rules; highlights: real Heading 1–6 styles, editable tables, Consolas code, native numbered lists and page-bottom footnotes, and native OMML equations for the supported LaTeX subset.
| Element | In the Word file |
|---|---|
| Headings H1–H6 | Native — Real Word Heading styles — they drive the Navigation Pane and auto-TOC |
| Bold / italic / strikethrough | Native — Word run formatting; inline code in Consolas |
| GFM tables | Native — Editable Word table: bold header row, cell borders |
| Links | Native — Clickable http/https/mailto hyperlinks |
| Bulleted lists, nested | Native — Nested up to 8 levels |
| Ordered (numbered) lists | Native — Real Word decimal numbering, nested up to 8 levels; each list restarts at 1 |
| Task lists (checkboxes) | Native — Real Word checkbox content controls — click to toggle in Word (engine v0.4.0) |
| Blockquotes | Native — Word “Quote” paragraph style — indented, italic, left rule (engine v0.4.0) |
| Fenced code blocks | Native — Consolas with a shaded background |
| Images | Approximated — data: PNG/JPEG images are embedded; remote/relative files become labelled placeholders — never fetched (privacy) |
| Footnotes | Native — Real Word footnotes — superscript reference markers, definitions at the bottom of the page |
| Math (LaTeX) | Native — Native editable equations (OMML) for the supported subset: fractions, roots, powers, bounds (\sum \int \lim), common symbols — the rest stays styled text, flagged per instance |
| Raw HTML blocks | Dropped — Removed for safety |
| Front-matter (YAML) | Native — title/author/description/subject/keywords map to Word core properties; other keys are listed in the report |
| Mermaid code blocks | Approximated — Rendered as code blocks; diagram rendering is planned |
Markdown → HTML
Exported with the Markdown to HTML tool. Keeps footnotes and sanitized raw HTML (opt-in) that the print exporters drop — a styled, self-contained page.
| Element | In the HTML |
|---|---|
| Headings H1–H6 | Native — Real <h1>–<h6> elements |
| Bold / italic / strikethrough | Native — <strong>, <em>, <del> |
| GFM tables | Native — Semantic <table> markup with alignment styles |
| Links | Native — <a href> with javascript:/data:/file: URLs stripped |
| Bulleted lists, nested | Native — <ul>/<ol> plus disabled checkbox inputs for tasks |
| Ordered (numbered) lists | Native — <ul>/<ol> plus disabled checkbox inputs for tasks |
| Task lists (checkboxes) | Native — <ul>/<ol> plus disabled checkbox inputs for tasks |
| Blockquotes | Native — <blockquote> elements |
| Fenced code blocks | Native — Fenced <pre><code class="language-…"> — highlighting is yours to style |
| Images | Native — Inline references kept exactly as written; nothing is fetched |
| Footnotes | Native — A real linked notes section at the end of the document |
| Math (LaTeX) | Native — Rendered as HTML math — real sub/superscripts, stacked fractions, overlined √ — for the supported subset (engine v0.4.0); anything else stays styled source text, flagged |
| Raw HTML blocks | Opt-in — Off by default; when enabled it runs through a strict sanitization allowlist |
| Front-matter (YAML) | Native — title/author/description map to the page <title> and meta description/author; other keys are listed in the report |
Bionic Reader — the bolding transform
Generated from the Bionic Reader tool's own capability table. The transform inserts bold markers into prose only; the five bionic exports reuse the PDF, Word, HTML and image engines above, so every element verdict in those sections applies to the bionic exports too.
| Aspect | Behavior |
|---|---|
| Word-prefix bolding | Native — Fixation strength 1–5 and saccade interval — prefix length scales with the word, minimum one letter |
| Headings, lists, tables, quotes | Native — Structure is parsed first — bolding applies to prose inside it, never to the structure itself |
| Code fences and inline code | Native — Excluded from bolding — code stays verbatim and copy-pasteable |
| Links and URLs | Native — URL targets are never bolded; only visible link text is — so every link still works and pastes cleanly |
| PDF / Word / HTML / Markdown export | Native — Real bold runs in .docx and selectable bold text in PDF; HTML is self-contained; Markdown keeps ** markers |
| Punctuation bolding | Opt-in — Optional — punctuation is merged into the same bold span, never creating malformed emphasis |
| CJK and space-less scripts | Dropped — Word-prefix bolding is undefined without spaces — segmentation is on the roadmap; those characters pass through untouched |
| EPUB in/out | Planned — On the roadmap — today paste text, Markdown, or load .md/.txt files |
What we do not support (yet)
- Math outside the supported subset, and Mermaid diagrams — Word equations (OMML) cover fractions, roots, powers, bounds and common symbols; everything else (environments, unusual macros) stays styled text, and Mermaid renders as code blocks. Both are engine updates in progress.
- PDF equations and per-page footnotes — PDF math is typeset with real sub/superscripts (v0.4.0) but not fully stacked fractions or editable OMML; footnotes collect into an end-of-document Notes section.
- Remote/relative images are never fetched — only
data:PNG/JPEG images embed into Word and PDF (privacy); the HTML exporter carries references as written. - Legacy .doc input — save as .docx first.
Every one of these limitations is also reported at conversion time — that is the point of the fidelity report.
Math support details
Since engine v0.3.0, the Word exporter renders the supported LaTeX subset as real, editable OMML equations: \frac, \sqrt[n]{}, powers and indices (^/_), \sum/\int/\lim with bounds, greek letters and the common operators, relations and arrows. Rejected-by-design: anything the translator cannot guarantee — environments like \begin{matrix}, custom macros, alignment characters — stays styled math text and is flagged per instance in the fidelity report. The PDF exporter typesets the same subset with real sub/superscripts, √ and ∑/∫ glyphs (no stacked fractions in PDF layout); the HTML exporter renders it with <sup>/<sub>, stacked CSS fractions and overlined roots. The translation is all-or-nothing per expression, so a native conversion never arrives half-converted.
Input limits
| Input | Limit |
|---|---|
| Pasted text | 2 MiB |
| DOCX / XLSX files | 10 MiB |
| Conversion timeout | 90 seconds with cancel |
| Everything else | No artificial caps — no daily quota, no signup, no watermarks |
Format support FAQ
Are my images uploaded when I convert?
No image ever leaves your device. As of engine v0.3.0, images supplied as data: URIs are embedded directly into the Word and PDF outputs — decoding happens locally. Remote and relative image files are still never fetched; they are listed in the fidelity report with placeholder treatment, and the HTML exporter keeps your original references verbatim.
Does Word output use real Word styles?
Yes. Headings map to native Heading 1–6 styles (they appear in Word's Navigation Pane and feed an auto-TOC), tables become editable Word tables, links stay clickable, and numbered lists and footnotes are real Word constructs. The fidelity report notes every element that maps only approximately.
Why does the fidelity score differ between Word and PDF for the same document?
The two engines are honestly different, not equally-marketed wrappers: math becomes native OMML equations in Word but stays styled text in the PDF, and footnotes are real Word footnotes while the PDF collects them into an end-of-document Notes section. The score reflects the direction you picked.
Is there a file-size or conversion limit?
There are no artificial caps. The workspace accepts text input up to 2 MiB and files up to 10 MiB, with a 90-second conversion ceiling per job — beyond that, the practical limit is your device's memory, because nothing runs on a server.