CatchTeXPDF table to LaTeX, without retyping
You need a results table from a paper in your own LaTeX document. Retyping twelve columns of numbers takes half an hour and invites typos. Here is how to get a clean booktabs table in a few seconds, and how the usual alternatives compare.
Updated 2026-10-05
The quick way: snip the table
- Open the PDF in CatchTeX (Ctrl+O). Using the built-in viewer matters: CatchTeX then reads the characters stored in the PDF instead of guessing them from pixels, so every number is exact.
- Drag a box around the table, including its header rows. Auto mode recognises it as a table; you can also pick the Table chip first.
- Check the preview. The result is compiled with the built-in pdfLaTeX and shown next to the code. Cells the reader is unsure about are highlighted, so you only check those.
- Copy LaTeX (Ctrl+Shift+C) and paste into Overleaf or your editor. You get a complete
tableenvironment with caption and label; add\usepackage{booktabs}to your preamble once.
The table is not in a PDF you can open, but in a slide, a web page or a scanned page? Press Ctrl+Alt+T and snip it from the screen. CatchTeX then uses offline OCR; numbers are usually right, but proofread them, as with any OCR.
What comes out
- booktabs rules (
\toprule,\midrule,\bottomrule) instead of a grid of\hlines. - Merged header cells as
\multicolumn. - Maths inside cells (Greek letters, ±, sub- and superscripts, arrows) as real LaTeX maths, not Unicode.
- Bold and italic values kept, so the best result stays marked.
Other ways, compared honestly
| Method | Exact numbers | Result | Your paper stays on your PC | Cost |
|---|---|---|---|---|
| Retype by hand | If you are careful | Whatever you write | Yes | Free, slow |
| Copy-paste from the PDF reader | Yes, but columns fall apart | A line of text to rebuild by hand | Yes | Free |
| Tabula / Excel, then Tables Generator | Yes for simple tables | LaTeX after 3 tools; merged headers and maths need fixing | Partly (web generator) | Free |
| ChatGPT or another AI chatbot | Not guaranteed: numbers can be changed silently | LaTeX, not compiled or checked | No, uploaded | Free tier / subscription |
| Mathpix Snip | Good OCR | LaTeX table | No, processed in the cloud | Subscription (comparison) |
| CatchTeX | Yes, read from the PDF itself | booktabs, compiled before you copy, doubtful cells marked | Yes, fully offline | Free edition; Pro once |
Tips for a clean result
- Include the header rows in your selection, and leave out the caption if you will write your own.
- Very wide tables: zoom out in the viewer so the whole table fits on screen before you snip.
- Table split across two pages: snip each part and join the rows in the code editor.
- Need
siunitxalignment on decimal points? Change the column spec toScolumns after pasting; the numbers are already plain. - Always look at the highlighted cells before you submit. CatchTeX tells you when it is unsure instead of guessing silently.
Questions
Does it work with scanned PDFs?
Yes, through offline OCR: a scanned page has no text layer to read, so CatchTeX recognises the image instead. Accuracy then depends on the scan quality, so check the numbers.
Do I need LaTeX installed?
No. CatchTeX includes its own pdfLaTeX for the preview. The table you copy works in Overleaf, TeX Live and MiKTeX.
Is it free?
The free edition has every mode with 10 snips a day (no limit during the public beta). Pro removes the limit with a one-time payment, no subscription.