Convert PDF to Text
A two-column page read as one column gives you the first line of column one, then the first line of column two, then back again. Layout recognition is what prevents that.
- Columns resolved during recognition, not after
- Works on scanned journal pages as well as born-digital
- Captions and footnotes stay with what they belong to
- Any output format: LaTeX, Word, Markdown or HTML
What goes wrong with two-column pages, and why
This is the most common complaint about OCR on academic papers, and it is a layout problem rather than a character-recognition one. Better character accuracy does not fix it.
Why most tools interleave the columns
A tool that reads a page line by line has no idea the page has columns. It crosses the gutter on every line, and the output is two documents shuffled together.
Layout is recognised, not guessed afterwards
v3/pdf and the Files API apply document-layout recognition automatically, so reading order is established as part of recognition rather than repaired later by heuristics.
It holds up on the awkward pages
A figure spanning both columns, a footnote at the bottom of one, a table that breaks the grid: these are where column detection by whitespace falls over.
What layout recognition covers
Reading order on multi-column pages
Not only two columns. Three-column layouts and mixed single- and multi-column pages in the same document are resolved per page.
Captions and footnotes, kept in place
Both are annotated structures in their own right, so they stay attached to the figure or page they belong to rather than landing mid-sentence.
Per-line geometry, so you can check it
include_line_data returns a box and a confidence score for every line, which is how you verify reading order on a long document instead of trusting it.
How to convert PDF to Text
1
Upload your PDF
Drag it into the Mathpix Snip web editor, or pick it from your files.
2
Convert it to Text
Recognition runs on the whole document: math, tables, figures and layout together.
3
Export or copy the Text
Download the Text, or copy it straight to the clipboard and paste it where you need it.
What Mathpix accepts, and the limits that apply
Which files you can send
PDF, EPUB, Word (DOCX, DOC), PowerPoint (PPTX), OpenDocument Text (ODT), DjVu, WordPerfect (WPD), Mobipocket (MOBI) and Amazon Kindle (AZW3)
What you get back
Mathpix Markdown, DOCX, LaTeX, HTML, Markdown, PDF, PPTX and XLSX, plus line-by-line JSON with per-line geometry.
The size limits that apply
One PDF, uploaded: 1 GB
Files API, per file: 15 GB
Files API, files per job: 200,000
Converting PDF to Text: common questions
Why do my columns come out interleaved?
Because the tool is reading the page as lines of text across its full width. It has no model of the page having columns, so it crosses the gutter on every line.
Do I need to set anything to get reading order?
No. v3/pdf and the Files API apply document-layout recognition automatically. enable_document_layout exists for the case where you post a full-page image straight to v3/text instead.
Does it work on scanned pages?
Yes. A scan goes through OCR and a born-digital page is read directly, and layout recognition applies to both.
What about three columns, or a mix?
Resolved per page, so a document that changes layout partway through is handled page by page rather than by one setting for the whole file.
Other formats you can convert PDF to
A whole PDF, out as LaTeX
The workflow for research papers.
Every output format from one upload
The hub for document conversion.
Line-by-line JSON with geometry
For checking reading order programmatically.