PDF / Image to Markdown

Drop a PDF, scan or screenshot and get editable Markdown right away — processed locally in your browser, never uploaded.

Short answer

Drop a PDF into the box above and you get Markdown in seconds. Text PDFs get headings from font size, keep lists, simple tables and monospace code blocks, and lose repeated headers/footers and broken line wraps. Scans and images go through basic in-browser OCR (tesseract.js): accuracy is limited and complex tables or layouts are not reconstructed. Preview side by side, copy, download .md or a ZIP — your files never leave your device.

Drop PDFs or images here, or click to choose PDF · JPG · PNG · WebP · multiple (up to 50) · nothing is uploaded
Options

Language data downloads on first use (about 1.6–2.9 MB), then the browser caches it.

Privacy: files are read and converted only in this browser (pdf.js and tesseract.js are self-hosted). 17NAS servers never receive your files or their text.

How to use

Four steps: add files → options → check → copy or download.

  1. Step 1

    Add files

    Drag PDFs, JPGs or PNGs into the box, or click “Choose files” — up to 50 at once.

  2. Step 2

    Adjust options

    Choose page markers and header/footer removal; for scans pick the OCR language (defaults to the page language).

  3. Step 3

    Check side by side

    Rendered preview on the left, editable Markdown on the right; step through pages of long files.

  4. Step 4

    Copy or download

    Copy in one click, download .md, or ZIP each page or all files.

Back to the tool

Key facts

Price
Free, no sign-up, no paid tier
Input
PDF; JPG, PNG, WebP, GIF, BMP images
Output
Markdown (CommonMark, GFM tables) as .md or ZIP
OCR
Basic OCR (tesseract.js): Simplified/Traditional Chinese, English, Japanese, Korean; limited accuracy
Privacy
Processed locally in the browser, no uploads
Limits
Up to 50 files; hundreds of text-PDF pages in seconds, OCR takes a few seconds per page
Browsers
Recent Chrome, Edge, Firefox, Safari (incl. mobile)
Last updated

Features and limits (measured)

Accuracy figures below are from our own tests on generated samples (character similarity, higher is better). Real files vary with fonts, image quality and layout.

ItemSupportNotes
Text in text-based PDFsYesRead directly; our English and Chinese samples (incl. CMap-encoded fonts) came out 100% intact
Headings, lists, code blocksYesHeadings from size/weight (up to 3 levels), bullet and numbered lists, monospace code blocks; all correct in our samples
Tables (text PDFs)PartialRebuilt as GFM tables from text positions; merged cells, tight columns or tables across pages may misalign
Headers/footers, line breaksYesRepeated headers, footers and page numbers removed; hyphenated words and wrapped lines rejoined
OCR on clean screenshotsPartialMeasured: English, Simplified/Traditional Chinese, Japanese ≈100%; Korean ≈87%
OCR on scans / small textPartialMeasured: noisy, slightly skewed scans ≈99%; 16 px text ≈98% (some words run together). Blurry, handwritten or low-res photos are much worse
Tables and layout in scansNoBasic OCR outputs paragraphs only; tables, columns and image positions are not reconstructed
Images, formulas, linksNoImages are not exported; math is output as plain text; PDF hyperlinks are dropped
Encrypted PDFsNoPDFs that need an open password cannot be processed
Multi-column layoutsPartialRead top to bottom; two-column papers may interleave

FAQ

Are my files uploaded?

No. PDF parsing and OCR run in your browser; this site only serves the code and language data and never receives your files or text.

How accurate is OCR on scans?

It is free, basic OCR (open-source tesseract.js). Clean printed text works well, but accuracy is limited: blur, handwriting and busy backgrounds cause errors, and tables are not rebuilt. Always proofread. There is no paid “high-accuracy” version.

Is my PDF text-based or a scan?

If you can select text with the mouse, it is text-based. In the default Auto mode, pages with text are extracted directly and only pages without text get OCR, so mixed files work too.

Are tables converted to Markdown tables?

Tables with clearly aligned columns in text PDFs become GFM tables; merged cells, multi-page or very dense tables may misalign. Tables in scans come out as plain text.

Why is OCR slow the first time?

The first run downloads the engine (about 4 MB) and language data (1.6–2.9 MB), which the browser then caches. Each page usually takes a few seconds, longer on phones.

Which OCR language for mixed text?

Pick the main language of the document; every option already includes English. It defaults to the page language — switch and click “Convert again”.

Can I convert in batch?

Yes, up to 50 files processed in order. View each file separately or click “All files ZIP” to download every .md at once.

Will big files freeze? Can I cancel?

Text PDFs are fast (60 pages in about 1 s in our test); OCR takes seconds per page. Cancel anytime — finished pages are kept and you can resume.

Can it convert password-protected PDFs?

PDFs that need an open password are not supported; remove the password in a PDF reader first.

How do I report a problem?

Email [email protected] with a description (no need to attach the file).

Feedback: [email protected]