PDF to CSV
What PDF to CSV actually does
A CSV where one line has one field and the next has four is not a CSV a spreadsheet or a script can load, and that is what a converter produces when it treats every visual line as a row. This workbench finds the recurring column alignment on a page, builds a rectangular grid from the rows that match it, and lists every line it kept out along with the reason. The grid is editable, and any excluded line can be pulled in as a row.
How to use it
- Choose a PDF up to 60 MB and 200 pages; the table is found locally.
- Pick a page, decide whether the first detected row is the header, and correct any cell directly in the grid.
- Add a column or a row where the detector was too strict, then download a CSV whose records all carry the same field count.
Useful for
- Pull a statement, price list, or results table out of a PDF into a spreadsheet or a script.
- Get a loadable CSV out of a report without a daily conversion cap or an upload.
- Fix a mis-split column once, in place, instead of re-editing the downloaded file.
Limits worth knowing
- Columns are detected from horizontal alignment, so a table whose cells wrap onto several lines produces one row per line.
- Merged cells and multi-row headers are flattened; short rows are padded with empty fields to keep every record the same width.
- Scanned tables carry no text and produce no grid. The OCR workbench comes first for those.
Questions people ask
Why do other PDF to CSV tools produce a file my spreadsheet rejects?
They emit one row per visual line, so titles and paragraphs become one-field records mixed in with the real table rows. RFC 4180 expects every record to carry the same field count, which is what this tool enforces.
What happens to lines that are not part of the table?
They are listed with the reason they were excluded, and each one can be added to the grid as a row if the detector was wrong.
Is the PDF uploaded?
No. Detection, editing and export all happen in this browser tab with locally bundled code.