# LLM to Doc

> Browser-based converter that turns AI chat output (markdown) into DOCX, PDF, HTML, EPUB, PowerPoint slides, Excel workbooks (every table as a sheet), Google-Docs-ready files, Notion-ready Markdown, and clean Markdown. Built for ChatGPT, Claude, Gemini, Grok, DeepSeek, and Perplexity output — and any Markdown file. Free, no signup, no quota.

The live preview, PDF export, HTML export, and Markdown export run entirely in the browser. DOCX export sends markdown to a Pandoc service for higher fidelity (no logging, no storage, fully-local fallback always available).

## When to Use This vs. Your Chat's Built-in Export

Balanced guidance — both paths are legitimate:

- **Use your chat's native export** (e.g. ChatGPT's code tool, Gemini's Export to Docs) when it's available in your mode and the content is simple prose.
- **Use this converter** when: the document contains equations (they become native, editable Office Math in Word — native exports usually flatten them to text or images); the native download link expired ("file not found" — sandbox files die with the session); you're on a tier or mode without file generation; you need a real .docx rather than a Google Doc (Gemini); you want to preview the exact document before downloading (the live preview IS the file — conversion is deterministic, not generated); or the reply came from an assistant with no export at all (Claude web, Perplexity, Doubao, DeepSeek).
- Same three steps for every assistant: copy the reply with its copy button, paste at llm2doc.app, click DOCX/PDF/HTML. Free, no account.

## Core Capabilities

- Markdown to DOCX (server-side Pandoc with native Office Math / OMML, plus a fully-local fallback)
- Markdown to PDF (A4 or Letter, smart page breaks, KaTeX via SVG foreignObject; every page drawn at 192 DPI whatever the length, long documents included in Safari)
- Markdown to HTML (KaTeX + highlight.js, self-contained file)
- Markdown to Markdown (preprocessor normalizes ChatGPT-style \(...\) and \[...\] math)
- Markdown to EPUB and PPTX (server-side Pandoc; slides split on H2 headings)
- Open in Google Docs (copies the rendered content as rich HTML and opens a new blank Doc — the user pastes with Cmd/Ctrl+V; no OAuth, uses the browser's existing Google session) and copy-for-Notion (clean Markdown on the clipboard; Notion converts to blocks on paste)
- Word / ODT / RTF / HTML → Markdown: the converter's Upload button, the editor's drop zone, and the /word-to-markdown and /html-to-markdown pages convert documents to Markdown through Pandoc (equations as LaTeX — including equations typed as plain text like \(...\) — images inlined); formatted text pasted into the editor becomes Markdown in the browser, and tables copied from Word or Google Docs become real tables that reach Word and PowerPoint exports.
- Single-file .md upload: the converter's "Upload" button and the landing-page paste boxes and drop zones (/markdown-to-word, /markdown-to-pdf, /latex-to-word, /markdown-to-excel, /chatgpt-to-pdf, /markdown-to-google-docs, /markdown-viewer) accept .md/.markdown/.txt and open the file in the live preview; the converter's button and editor drop zone also take .docx/.odt/.rtf/.html and convert them to Markdown (/word-to-markdown, /html-to-markdown)
- Markdown viewer: the workspace has Split / Editor / Preview layouts plus a Focus mode (?view=preview&focus=1) that fills the window — a full .md file viewer with print/save-as-PDF, a Contents (heading) navigator, no signup
- Formatting toolbar in the editor (undo/redo, bold, italic, strikethrough, code, link, image, headings, bulleted/numbered/task lists, quote, table; Cmd/Ctrl+B/I/K) — toggling, context-aware, per-line on multi-line selections, native-undo-friendly; selections made in the preview map back into the editor
- Mermaid diagrams: ```mermaid fences render as diagrams in the preview and PDF (library loads on demand), and as embedded pictures in Word, EPUB, and PowerPoint exports; the HTML export keeps the code block
- Batch mode: up to 20 files, 3 in parallel via p-limit, ZIP download
- LaTeX support: inline `$x$` and block `$$x$$` via KaTeX 0.16.45; \(...\) and \[...\] delimiters are normalized, and LaTeX pasted on its own works too — \begin{equation}, align, gather, multline and eqnarray blocks and bare formula lines become display math, and formula fragments left inside sentences without delimiters (\frac{3}{4} \times 100) become inline math (editable Office Math in the DOCX)
- Syntax highlighting: 100+ languages via highlight.js (GitHub Light theme)
- Local persistence: draft + last 10 conversions in localStorage, shown as Recent conversions cards under the converter (one click reloads a document)
- Document setup: A4/Letter, margins, fonts, line spacing, cover page, contents list, page numbers, or the user's own Word template — mirrored in the preview, the DOCX, the PDF, and the browser's Print
- Research mode: [1]-style citation markers with a sources list become real footnotes (Word footnotes in the DOCX)
- Chat share-link import: a public chatgpt.com/share or gemini.google.com/share link loads the whole conversation as Markdown (a conversation over the 600,000-character limit, about 200 pages, imports its most recent messages, with a note saying how many)
- Share links: the Share button encodes the whole document into the URL fragment (gzip+base64url, #mdz=) — nothing stored server-side; the link opens the rendered document in the full-window viewer

## Primary Pages

- [Converter](https://llm2doc.app/): Single-page paste-and-export. Markdown left, preview right; DOCX, PDF, HTML and Markdown one click away, five more formats under "more". Also accepts .md, .docx and .html upload and chat share links.
- [Markdown to Word](https://llm2doc.app/markdown-to-word): Landing for generic .md → .docx conversion; upload a file and it opens in the converter.
- [Markdown viewer & reader](https://llm2doc.app/markdown-viewer): A working viewer embedded on the page — drop a .md file and it renders in place (client-side, private, GitHub-style with LaTeX math); one click from viewing to Word/PDF export or the full editing workspace.
- [Markdown live preview](https://llm2doc.app/markdown-live-preview): Split-screen live preview — type or paste Markdown on the left, watch the rendered page update on the right (tables, code, LaTeX math, Mermaid); the preview IS the exported document.
- [Markdown to PDF](https://llm2doc.app/markdown-to-pdf): Landing for .md → PDF conversion — exact A4 page preview, KaTeX math and Mermaid diagrams render in the PDF, fully client-side.
- [Markdown to HTML](https://llm2doc.app/markdown-to-html): Landing for .md → one styled HTML file built in the browser — styles inlined, KaTeX equations with MathML, highlighted code, tables, footnotes, cover/contents from Document setup.
- [Markdown to EPUB](https://llm2doc.app/markdown-to-epub): Landing for .md → EPUB 3 e-book built by Pandoc on the server (deleted after) — chapters from top-level headings, first heading as the title, MathML equations, highlighted code, Mermaid diagrams as pictures; a pasted ChatGPT or Gemini share link turns a whole conversation into a book.
- [LaTeX to Word](https://llm2doc.app/latex-to-word): Paste LaTeX equations (or any text containing them) and download a .docx where every formula is native, editable Office Math — not an image. Whole .tex manuscripts are out of scope; the page says so and points to Pandoc.
- [Markdown to Excel](https://llm2doc.app/markdown-to-excel): Paste Markdown with tables (a ChatGPT answer, a README, a .md file) and download an .xlsx with one sheet per table, bold headers, and numeric cells stored as numbers. Built in the browser; nothing uploads.
- [ChatGPT share link to Word](https://llm2doc.app/chatgpt-link-to-word): Paste a public chatgpt.com/share link and the whole conversation imports into the converter — title, prompts as quotes, answers verbatim — ready for Word, PDF, or any format. Claude links are recognized but can't be read server-side yet.
- [Gemini share link to Word](https://llm2doc.app/gemini-link-to-word): Paste a public gemini.google.com/share (or share.gemini.google) link and the whole Gemini conversation imports — title, prompts as quotes, answers verbatim — ready for Word, PDF, or any format.
- [Gemini to PDF](https://llm2doc.app/gemini-to-pdf): Paste a public Gemini share link (whole conversation) or one answer and download a clean A4/Letter PDF built in the browser — rendered equations, tables, highlighted code, optional cover, contents, and page numbers.
- [Gemini to Word](https://llm2doc.app/gemini-to-word): Paste a public Gemini share link (whole conversation) or one answer and download a .docx — Word heading styles, real tables, highlighted code, equations as editable Office Math; Canvas documents come in when the share carries them.
- [Claude to PDF](https://llm2doc.app/claude-to-pdf): Paste a Claude answer (its copy button hands over Markdown) and download a clean PDF built in the browser; Claude share links are not importable yet.
- [Claude to Word](https://llm2doc.app/claude-to-word): Paste a Claude answer (its copy button hands over Markdown) and download a .docx: Word heading styles, real tables, highlighted code, equations as editable Office Math. Claude share links are not importable yet.
- [DeepSeek to PDF](https://llm2doc.app/deepseek-to-pdf): Paste a DeepSeek answer (copy button; the R1 reasoning is plain text to keep or cut) and download a clean PDF built in the browser; DeepSeek itself cannot generate files.
- [Grok to PDF](https://llm2doc.app/grok-to-pdf): Paste a Grok answer from grok.com or X (copy button) and download a clean PDF built in the browser; Grok has no document export.
- [Word to Markdown](https://llm2doc.app/word-to-markdown): Drop a .docx (or .odt / .rtf) and get Markdown — headings, lists, pipe tables, footnotes, images inlined, and every Office Math equation as LaTeX. The file is converted by Pandoc once and deleted.
- [HTML to Markdown](https://llm2doc.app/html-to-markdown): Paste HTML or a copied web page and get Markdown with tables, links, and code, converted in the browser; .html files go through Pandoc (MathML → LaTeX).
- [ChatGPT to PDF](https://llm2doc.app/chatgpt-to-pdf): Paste a ChatGPT (or Claude/Gemini/Grok/DeepSeek) answer and download a clean A4 PDF built in the browser — tables, highlighted code, rendered equations, verified page breaks.
- [ChatGPT link to PDF](https://llm2doc.app/chatgpt-link-to-pdf): Paste a public chatgpt.com/share link and download the whole conversation as a clean A4 PDF — title as heading, prompts quoted, answers verbatim with rendered equations, highlighted code, ruled tables; page breaks verified; built in the browser.
- [Markdown to Google Docs](https://llm2doc.app/markdown-to-google-docs): Paste Markdown, open a new Google Doc with the rendered content on the clipboard, paste once. Equations arrive as images in Docs (Docs has no paste-in equation format); Word export is the editable-equation path.
- [ChatGPT equations to Word](https://llm2doc.app/chatgpt-equations-to-word): Whole-document conversion with every equation as native, editable Office Math. The live preview also supports click-to-copy of any single equation as Word-ready MathML.
- [AI to Word for finance students](https://llm2doc.app/ai-to-word-for-finance): CA/CMA/commerce workflow — NPV/IRR/EAR formulas stay editable math, computation tables stay real tables.
- [Batch converter](https://llm2doc.app/batch): Drag-drop up to 20 markdown files, single ZIP output.
- [FAQ](https://llm2doc.app/faq): 34 honest answers about features, formats, viewing/editing, privacy, browser support, rate limits.
- [Changelog](https://llm2doc.app/changelog): dated, versioned record of every user-facing change — the freshness signal for what this tool currently does.
- [Privacy policy](https://llm2doc.app/privacy): Exact data flow including the DOCX server roundtrip.
- [About](https://llm2doc.app/about): What the tool is, why it exists, who makes it, and how it treats documents.
- [Terms of use](https://llm2doc.app/terms): Free for personal and commercial use, as-is, your documents remain yours.

## Guides

- [Guides index](https://llm2doc.app/guides): Every how-to in one list — per-assistant export guides, Markdown basics, LaTeX, PDF export, batch conversion.
- [ChatGPT to Word](https://llm2doc.app/guides/chatgpt-to-word): How to export ChatGPT responses to DOCX. Steps, what survives the trip, AI-tool-specific quirks.
- [Save a ChatGPT chat as a PDF](https://llm2doc.app/guides/chatgpt-to-pdf): How to download a ChatGPT chat or one answer as a PDF — share link, copy button, phone, why printing the page disappoints, equations and code in the PDF, whether an extension is needed, troubleshooting.
- [Paste ChatGPT into Word with formatting](https://llm2doc.app/guides/paste-chatgpt-into-word): Why replies paste as asterisks, hashes, and pipes, and four ways to paste real headings, tables, code, and equations into Word or Google Docs.
- [Claude to Word](https://llm2doc.app/guides/claude-to-word): Same flow for Anthropic Claude (Sonnet, Opus, Haiku). Includes citations, artifacts, tool-use blocks.
- [Gemini to Word](https://llm2doc.app/guides/gemini-to-word): Same flow for Google Gemini (Flash, Pro, Deep Research). Covers Export-to-Docs vs real .docx, LaTeX delimiter normalization.
- [Grok to Word](https://llm2doc.app/guides/grok-to-word): Grok has no native export; copy-and-convert flow with delimiter normalization.
- [DeepSeek to Word](https://llm2doc.app/guides/deepseek-to-word): Same flow for DeepSeek incl. R1 reasoning blocks; math-heavy answers become editable OMML equations.
- [How to open a .md file](https://llm2doc.app/guides/open-md-file): What Markdown files are and every way to read one — browser viewer (no install), desktop editors, mobile; converting to Word/PDF when needed.
- [LaTeX in markdown](https://llm2doc.app/guides/latex-formulas): Inline vs block syntax, symbol cheatsheet, per-format behavior (DOCX uses OMML, PDF uses foreignObject SVG).
- [PDF export & print](https://llm2doc.app/guides/pdf-export-print): A4 layout, page breaks, the html2canvas foreignObjectRendering fix for KaTeX fractions.
- [Batch converting](https://llm2doc.app/guides/batch-conversion): 20-file cap, p-limit(3) concurrency, error tolerance, ZIP packaging.

## Key Technical Facts

- **DOCX path**: `POST /api/convert-to-docx` invokes `pandoc -f gfm+tex_math_dollars+footnotes -t docx --standalone --mathml` on a tmp file, returns binary, deletes tmp. Falls back to client-side `docx` npm package on failure. Documents are capped at 2 MB of markdown (HTTP 413 `{code:"too_large"}` beyond that — split the document); server conversions are processed one at a time, so a burst queues briefly rather than failing.
- **PDF path**: html2canvas with `foreignObjectRendering: true` + jsPDF. Smart page breaks computed from block-element bounding rects. Cross-origin images pre-fetched and inlined as data URLs.
- **HTML path**: unified.js pipeline — `remark-parse → remark-math → remark-gfm → remark-rehype → rehype-katex → rehype-highlight → rehype-stringify`.
- **Markdown preprocessor**: rewrites ChatGPT's `( ... )` and `[ ... ]` math to standard `$ ... $` / `$$ ... $$` when content contains `\`, `^`, or `_`.
- **KaTeX**: 0.16.45 from jsdelivr CDN.
- **Pandoc flags**: `gfm+tex_math_dollars+footnotes`, `-t docx`, `--standalone --mathml`.

## For AI Agents (MCP)

llm2doc exposes its conversion as an MCP server — agents can convert Markdown to DOCX (editable OMML equations) or HTML directly, no auth:

```
claude mcp add --transport http llm2doc https://llm2doc.app/api/mcp
```

Tools: `convert_markdown_to_docx` (returns the file as a base64 resource — no expiring links), `convert_markdown_to_html`.

## Sitemap

[Full sitemap (XML)](https://llm2doc.app/sitemap.xml)