PDF to Markdown
LocalConvert PDF to Markdown in your browser — H1–H3 headings, numbered lists and pipe tables kept. Free, no upload, no sign-in; optional AI summary with your key.
Drag files here, or click to browse
Supports PDF
Choose fileWhat is it? PDF to Markdown converts a PDF text layer into Markdown — headings ranked H1–H3 by font size/weight, numbered lists keep their numbers, aligned rows become pipe tables — useful for documentation and AI pipelines. An optional AI summary/translation (your own LLM key, browser-direct to the third-party endpoint) works on the result.
TL;DR: Drop in a PDF and its text layer becomes Markdown: headings ranked H1–H3 by font size and weight, numbered lists keeping their numbers, aligned rows becoming pipe tables. Free, local, no upload; the optional AI summary/translation uses your own API key.
How to use
- Open a PDF — extraction runs page by page and can be canceled at any time.
- Review the Markdown.
- Copy or download the .md.
- Optionally expand "AI summary / translation" at the bottom: enter your own LLM API key and endpoint to summarize the result or translate it into 10 languages — requests go from your browser straight to that provider.
FAQ
Are images included?
No — only the text layer is converted.
How are headings and tables detected?
Heuristically: heading levels come from font size relative to the body plus bold weight; aligned rows become pipe tables only when confidence is high — low-confidence candidates are kept as plain text and disclosed.
Can it summarize or translate the document?
Yes — the optional AI panel at the bottom of the page offers a summary (abstract + key points) or translation into 10 languages. It uses your own API key and sends the selected text directly from your browser to the LLM endpoint you configure (default api.openai.com); it does not pass through Vecaria.
Is it local?
Conversion is local — the PDF and its Markdown never touch Vecaria servers. The optional AI summary/translation is different: it sends the selected text directly from your browser to the third-party LLM endpoint you configure (your own API key; it does not pass through Vecaria). Skip it and everything stays local.
Does it work on scanned PDFs?
No — the tool converts the text layer, so a scan without one produces nothing useful; the page flags it. Run PDF OCR first, then convert the text.
Are equations converted to LaTeX?
No — math is not recognized. The text layer is converted as plain text, so equations come through as whatever characters the PDF stores. LaTeX-grade math extraction needs an OCR service built for it, which is a different tool category.
Is the output good for LLMs or RAG pipelines?
Yes — that is the main use case. Structured Markdown with ranked headings, numbered lists and pipe tables is token-efficient and reads natively by LLMs, docs systems and knowledge bases.
Is it free? Do I need an account?
Single-file conversions use a free daily allowance with no sign-in — the quota resets every 24 hours. Everything runs in your browser and the PDF is never uploaded.
Examples
- Drop in a PDF → converts to Markdown automatically → download the .md (or AI-summarize/translate it with your own key)
Supported formats
Limitations
- Scanned PDFs need OCR first.
- Heading levels (H1–H3) and tables are heuristic; low-confidence table candidates keep the original text lines (disclosed in the page).
- Complex layouts may need manual cleanup.
- Optional AI summary/translation (BYO key) sends the selected text from your browser straight to the third-party LLM endpoint you configure (default api.openai.com); it does not pass through Vecaria — local conversion itself uploads nothing.
Privacy
Privacy first: this tool runs entirely in your browser — your input and files never leave your device, and everything is cleared when you close the page.
Related tools
Standards & comparison
| Dimension | Vecaria (this tool) | Online converters (cloud) | Desktop professional CAD |
|---|---|---|---|
| Privacy | Runs locally in your browser; the file is not uploaded | You upload the file to a third-party server | Runs locally |
| Wait | No queue — as fast as your own device | Upload plus cloud queue | Local application startup |
| Standards support | PDF (ISO 32000) | Varies by service, often older subsets | Full vendor protocol sets |
| Install footprint | 0 MB — runs in the browser | 0 MB | Several GB of storage |