PDF to CSV converter: tables to clean, quoted CSV

Drop a PDF and download its tables as a CSV file: the original column headers once at the top, one line per row, and fields with commas or quotes escaped so every row has the same number of columns. PDFs exported from software and image-only scans both work.

Two real PDFs, start to finish

We made two PDFs that are awkward to turn into CSV: a two-page supplier price list whose descriptions are full of commas and inch marks (1/2") and whose part numbers start with 0, and a delivery note that is only a picture of a page, with no text to select. Each went through this page with "All tables", and we downloaded the CSV. Every cell was compared with the table we typed, and the file was parsed back with a standard CSV reader. Nothing below was edited.

1. A two-page price list with commas and quotes

First page of a plumbing supply price list with part number, description, unit and price columns; descriptions contain commas and inch marks — Input, page 1 of 2: the table continues on page 2 under a repeated header.

ExtractFox result with all 34 price list rows in one table — What this page showed: 136 of 136 cells right, one table.

The CSV has one header and 34 rows, and every row parses to exactly 4 fields. A description such as Copper pipe 1/2", type L, 10 ft is wrapped in quotes with the inch mark doubled, so its commas stay inside the field. Part numbers stay 004512, and 1,249.00 is written as 1249 without the thousands comma that would otherwise split the column.

Part no.,Description,Unit,Price (USD)
004512,"Copper pipe 1/2"", type L, 10 ft",ea,1249
004513,"Copper pipe 3/4"", type L, 10 ft",ea,1874.5
004610,"Elbow 90°, 1/2"" copper",bag 25,38.75

First lines of the downloaded CSV.

2. A scanned delivery note

Image-only PDF of a medical supply delivery note with line, item code, description, lot, ordered and delivered columns — Input: an image-only PDF we made from a rendered page (0 characters of selectable text).

ExtractFox result table with eight delivery lines, lot numbers and ordered and delivered quantities — 48 of 48 cells right.

Lot numbers like L24-0918 and S24-1103 are where plain OCR slips: Tesseract read them as 124-0918, $24-1103 and 624-0930. In the CSV every lot is right, and Ordered and Delivered are plain numbers, so the back-ordered line (30 ordered, 0 delivered) is easy to filter.

Results side by side

Cells right = the value in the right row and column, compared with the table we typed. pdftotext and Tesseract return lines of text with no columns, so for them we counted how many of the true values appear anywhere in their output, which flatters them.

PDF	Cells	ExtractFox: cells right	CSV rows / fields per row	Seconds	Text tool: values found
Price list (text layer, 2 pages)	136	136	34 / always 4	6.4	pdftotext: 136
Delivery note (image only)	48	48	8 / always 6	4.7	Tesseract: 43

Two PDFs is a demonstration, not a benchmark. Both were legible and laid out as plain grids. Where it gets harder: low-resolution scans, headers that span two rows, and very long statements. A cell that can't be read comes back empty rather than guessed.

Why extracting the text isn't a CSV yet

pdftotext found every value in the price list, because the PDF has a text layer. It gives lines with columns lined up by spaces, and the description column itself contains spaces and commas, so splitting on either breaks rows. Saving that text as .csv gives one field per line. The scanned delivery note has no text layer, and Tesseract's lines mix up letters and digits in exactly the codes you would join on.

Line Item code Description Lot Ordered Delivered
1 AX-3301 Nitrile gloves, size M (100) 124-0918 40 40
2 AX-3302 Nitrile gloves, size L (100) 124-0918 40 36
3 BX-1150 Face masks, type IIR (50) M24-0702 25 25
4 Cx-0420 Hand sanitizer gel 500 ml $24-1103 60 60

Tesseract on the delivery note, first table lines.

What the CSV file looks like

Comma-separated, CRLF line ends, one header row. A field with a comma, quote or line break is wrapped in double quotes and inner quotes are doubled (RFC 4180).
UTF-8 with a byte-order mark, which is how Excel knows to show 90° and not 90Â° when you double-click the file. In Python, read it with encoding="utf-8-sig" so the first header has no stray character; pandas.read_csv(path, encoding="utf-8-sig").
Amounts and quantities are plain numbers (1249, -114.9): no currency symbols, no thousands separators.
A table that continues on the next page becomes one block of rows with one header. When the PDF has several tables, each row carries the table's caption and page so you can tell them apart.
A value that starts with =, +, - or @ and is text gets a leading apostrophe, so a spreadsheet doesn't run it as a formula.

Leading zeros and Excel

The CSV keeps 004512 as written. Excel, when it opens a CSV by double-click, decides on its own that 004512 is a number and shows 4512. If the codes matter, either download .xlsx from PDF to Excel, where codes are stored as text, or import the CSV with Data > From Text/CSV and set that column to Text before loading. In Google Sheets, File > Import with "Convert text to numbers, dates, and formulas" unticked keeps them too.

Getting the CSV you need

"Everything" (every table) keeps the PDF's own column headers. For fixed columns across many different PDFs, describe them in the box: "date, description, amount, balance" gives the same header every time.
For many PDFs, select them together (up to 25 per batch with a plan or a credit pack): one CSV with a source_file column.
From code, the API returns the same extraction as JSON, ready to write out as CSV.

Sources

pdftotext 26.10.0 (Poppler) run as pdftotext -layout file.pdf -, October 14, 2026
Tesseract 5.5.3 run as tesseract page.png - --psm 6 on the scanned page, October 14, 2026
RFC 4180, Common Format and MIME Type for Comma-Separated Values (CSV) Files
ExtractFox runs on the same PDFs, October 14, 2026; script in the repo at scripts/samples/pdf-tables.mjs, method on how we test

Frequently asked questions

How do I convert a PDF to CSV?+

Upload the PDF on this page, keep "Everything" selected, check the result on screen, then click CSV.

What happens if a field contains a comma, quote or line break?+

It is wrapped in double quotes and any quote inside it is doubled, as RFC 4180 describes, so Copper pipe 1/2", type L stays one field. Every row has the same number of fields.

Will the CSV open correctly in Excel and Google Sheets?+

Yes. The file is UTF-8 with a byte-order mark, which is what makes Excel show 90° rather than 90Â° when you double-click it. Google Sheets reads it with File > Import. One catch: Excel turns codes like 004512 into 4512 when it opens any CSV; download .xlsx from the PDF to Excel page instead, or import with Data > From Text/CSV and set that column to Text.

Does it work on scanned PDFs?+

Yes. OCR is built in. In our test an image-only delivery note came back with 48 of 48 cells right, including lot codes like L24-0918 that plain OCR misread.

What about tables that span several pages?+

They are joined into one block of rows with a single header; the header repeated on each page is dropped.

What delimiter and encoding does it use?+

Comma delimiter, CRLF line ends, UTF-8 with a byte-order mark, one header row. In Python open it with encoding="utf-8-sig" so the first column name has no stray character; most database loaders accept it once the BOM is skipped or stripped.

Are numbers formatted for import?+

Amounts and quantities are plain numbers like 1249 or -114.9: no currency symbols, no thousands separators, and a value in parentheses becomes negative.

Can I get the same columns from many different PDFs?+

Yes. Instead of keeping "Everything", describe the columns, for example "date, description, amount, balance", and every file comes back with that header. Up to 25 PDFs per batch with a plan or credit pack, in one CSV with a source_file column.