Convert PDF files to Markdown, without uploading them anywhere
A file-to-Markdown converter that runs entirely inside your browser. Nothing is sent to a server, and PDF tables are rebuilt from their coordinates instead of collapsing into one column.
Drop a file here
or click to choose one from your device
Choose a filePDF is inspected for a text layer before anything else happens.
Result
Built for the two things people actually complain about
Read the threads about PDF-to-Markdown tools and the same two problems come up: the file has to be uploaded, and the tables arrive broken. This tool exists to answer both.
Never uploaded
The conversion runs on your own device. There is no server holding your document and no upload wait before you see output.
Tables survive
Every text fragment keeps its x and y position, so rows and columns are reconstructed. A plain text extractor throws that away — that is why its tables collapse.
Scans are caught first
An image-only PDF has no text to extract. The byte stream is checked for font objects, and you are told before you waste time on a conversion.
No signup, no quota
No account, no email, no daily limit. Open the page and convert.
Real Markdown, not text
Headings, lists and tables come out as Markdown syntax, ready to paste into a note, a repo or an AI prompt.
Nothing is stored
There is nowhere for your file to be kept, because it was never sent. Close the tab and nothing remains.
How the conversion works
Four steps, all on your machine.
The browser reads the file locally. No network request ever carries it.
Font objects are counted in the byte stream. No fonts means a scan, and a scan needs OCR — not a converter.
Fragments sharing a baseline become one row; their x order becomes the columns. This is the step most converters skip.
Nothing is queued and nothing is kept on any server.
What a plain text extractor does to your table
Measured on the same one-page PDF, on one machine.
| Extraction method | What came out |
|---|---|
| Plain text extraction | Department / R&D / Marketing / Amount / 1200 / 800 — row pairing lost |
| Coordinate-based reconstruction | Department | Amount / R&D | 1200 / Marketing | 800 — intact |
It is not a smarter model. It is the same text plus the position of every word. See the full set of measurements.
Supported formats
What converts today, and what does not.
| Format | Status |
|---|---|
| PDF (with a text layer) | Converts in the browser, tables reconstructed from coordinates |
| PDF (scanned / image only) | Detected and reported — this file needs OCR |
| Markdown, plain text, logs | Converts as-is |
| HTML | Headings, paragraphs and tables preserved |
| CSV | Converted to a Markdown table |
| JSON | Wrapped as a fenced code block |
| Word, Excel, PowerPoint, EPUB | Not enabled yet — on the roadmap |
Questions
Short answers, no marketing.
Are my files uploaded to a server?
No. The conversion runs inside your browser. Your document is read from your disk, processed locally, and never transmitted.
Does it cost anything?
No. There is no account, no quota, no paid tier and no watermark.
Why do PDF tables come out broken elsewhere?
Most converters are plain text extractors — they read a PDF as a stream of words and discard where each word sat. Columns then stack into a single list. Keeping the coordinates is what allows the table to be rebuilt, and that is what this page does.
Is the conversion always perfect?
No. Simple tables rebuild cleanly. Multi-column magazine layouts, merged cells and tables drawn as images can still come out wrong. Anything that was a picture cannot be recovered as text at all.
Why does my scanned PDF return nothing?
A scan is a picture of a page. There is no text in it, so no converter can extract any — it needs OCR. This page detects that case and tells you instead of handing back an empty file.
Do I need to install anything?
No. The first PDF you convert downloads a small engine once (about 8 MB) and then caches it. Text formats need nothing at all.
Is this affiliated with Microsoft?
No. MarkItDown is an open-source Microsoft project, and this is an independent tool built around the same idea. It is not affiliated with or endorsed by Microsoft.
What should I do with the Markdown afterwards?
It is plain text, so it drops into a notes app, a Git repository, a documentation site, or an AI prompt. Markdown is usually easier for a model to work with than a PDF.
Convert a file now
Nothing leaves your device.