Tabulyn

Convert PDF tables to CSV

Get plain, portable CSV out of a PDF — the format that loads cleanly into pandas, R, Postgres, or any script you already have. Review the extracted rows first, then download.

Drop your PDF here, or browse files

PDF · up to 10 MB · up to 30 pages

  • No signup needed
  • Auto table detection
  • Files auto-deleted

What the output looks like

One header row, then one line per data row, UTF-8 encoded with standard CSV quoting:

Account,Debit,Credit,Balance
Cash,"$5,000.00",$0.00,"$5,000.00"
Revenue,$0.00,"$12,300.00","$12,300.00"

Values containing commas are quoted, so the file parses correctly with a default reader — no custom delimiter handling needed.

CSV or Excel?

Choose CSV when

  • You're loading it into code or a database
  • You want a diff-friendly, plain-text file
  • Another tool expects a simple import format

Choose Excel when

  • You need numbers and dates already typed
  • You want several tables as separate worksheets
  • You'll build formulas on top of the data
Read the Excel guide →

Questions about CSV export

What do I get when the PDF has more than one table?

A single CSV can only hold one table, so when you select several, Tabulyn packages them as a ZIP containing one CSV per table, each named for the page and table it came from. Select a single table and you get a plain .csv straight back.

How is the CSV encoded and quoted?

UTF-8, comma-separated, with standard quoting: any value containing a comma, quote, or newline is wrapped in double quotes and internal quotes are doubled. That is the format Excel, pandas, and Postgres COPY all expect by default.

Are currency symbols and thousands separators stripped?

No — CSV output preserves the value as displayed, so $1,500.00 stays $1,500.00. CSV has no type system, so stripping formatting would be a lossy guess. Parse it in your own pipeline, or use the Excel export where values are written as real typed numbers.

Can I edit rows before exporting?

Yes. Every detected table opens in an editable preview first — fix a mis-read cell, deselect tables you don't need, then export. The CSV reflects exactly what you see.