Search DevTools

Jump to any tool or page

PDF to Markdown & JSON

Extract a PDF's text as Markdown, or as a structured JSON report with page counts, layout analysis and OCR hints. The PDF is parsed in your browser.

Source

no file yet

Drop PDF here or click to browse

Supports files up to 50 MB

Markdown profile

Fidelity keeps the source layout. Compact strips it down for fewer tokens.

For example 1, 4, 8-12. Leave empty for the whole document.

Markdown

Choose a PDF or enter a downloadable link to view extracted content.

Data Transformation

About PDF to Markdown & JSON

Extract text, metadata, and structure from a PDF so it can be searched, diffed, or fed into a pipeline. PDF is a page-description format rather than a document format: it records which glyphs are painted at which coordinates, with no inherent notion of a word, a line, or a paragraph, so every extractor is reconstructing structure that was never stored.

Related Data Transformation tools

JSON FormatterImage to JSONPII RedactorSheet to JSONJSON to CSV/XLSX

Frequently asked questions

Why does extracted text come out with words jammed together or split apart?
PDFs position text runs by coordinate, and spaces are frequently not encoded as characters at all — the renderer simply moves the cursor. Extractors infer word boundaries from horizontal gaps, so tight kerning produces runtogetherwords while justified text with stretched spacing produces s p l i t ones. Ligatures compound the problem: a single fi glyph may extract as one codepoint rather than two letters, quietly breaking substring searches for words like "file" or "confirm".
Why does a scanned PDF return no text at all?
Because there is none. A scan is a raster image wrapped in a PDF container, so there are no glyph operators to extract — the file is a photograph of a page. That requires optical character recognition, which is a separate process from extraction. Some scans are hybrids: an image layer with an invisible OCR text layer behind it, produced by scanner software. Those extract fine, but inherit the OCR's errors, and the extracted text may not match what a human reads on the page.
How does column and table extraction go wrong?
Reading order in a PDF is the order operators appear in the content stream, which usually reflects the generating tool's internal sequence rather than human reading order. Two-column layouts often extract as interleaved lines, alternating between columns. Tables are worse: cell boundaries are drawn as separate vector lines with no relationship to the text objects, so a table extracts as a stream of values with no row or column association. Position-aware extraction that clusters by coordinate recovers more than plain stream reading.
What do the encryption and permission flags mean for extraction?
PDFs support two distinct protections. A user password encrypts the content, and without it nothing can be read. An owner password leaves the file readable but sets permission flags — no printing, no copying — which are advisory: the content is decrypted with an empty user password, and honouring the flags is a convention libraries choose to follow rather than a cryptographic barrier. So a file that refuses copy-paste in a viewer may still extract cleanly, which is a licensing question rather than a technical one.
What metadata does a PDF carry beyond the visible text?
Rather a lot, and it is a routine leak. The document info dictionary and XMP packet typically record author name, the producing software and version, and creation and modification timestamps. Files converted from Word often retain the original author and company. Beyond that, deleted content can survive: incremental updates append changes without removing prior versions, so earlier revisions may remain recoverable in the byte stream, and black rectangles drawn over text redact nothing at all.