Image to CSV - Extract Tables From Photos & Screenshots

Yes — converting a table image to CSV is free, needs no account, and the downloaded file carries no watermark or branding. Images are processed on our servers and deleted automatically after the CSV is generated.

Upload the image, the OCR detects the table structure, and a UTF-8 encoded CSV is generated and downloaded. Commas inside cells are quoted, headers stay in the first row, and trailing whitespace is stripped, so the file is ready to drop into Google Sheets, Excel, Numbers, or any BI tool that consumes CSV.

Yes. The CSV is RFC-4180 compliant with UTF-8 encoding and CRLF line endings, which Google Sheets, Excel, Numbers, Airtable, Notion and most CRMs ingest without prompting for advanced import options. Just choose File → Import and pick the downloaded .csv — no extra cleanup needed.

Yes. Currency markers ($, €, £, ¥, ₹), percentage signs and common date formats (YYYY-MM-DD, MM/DD/YYYY, DD-MM-YYYY) are preserved verbatim inside cells. Re-format inside your spreadsheet of choice if you want them treated as numbers — most apps detect the patterns automatically and convert.

The default separator is a comma and the encoding is UTF-8 (BOM optional), which suits English, Spanish, French and most Western datasets. For locales where decimal commas are common (Germany, France), open the CSV with the semicolon import option in Excel, or use the CSV-to-Excel converter in our toolbox for a styled .xlsx instead.

Yes. If the photo contains several distinct tables, the tool detects each one and lets you download a ZIP of separate CSV files — one per table — so you can drop them into different sheets or import them as separate datasets into your CRM, accounting tool or BI dashboard.

Uploads run on isolated servers, are deleted within minutes, and never train any model. No watermark, no attribution, no tracking — the CSV is yours to use commercially. That makes the tool safe for sales pipelines, customer lists, inventory exports and any data you would not want stored long-term.

The Default engine (PaddleOCR PP-Structure) is tuned specifically for table structure and is the best starting point for most screenshots and photos of tables. Table AI (beta) is an experimental alternative worth trying if Default misreads a complex or borderless table; Engine 1 and Engine 2 are general-purpose OCR fallbacks useful when the table engines struggle with unusual fonts or handwriting.

Open the Language picker in the Settings panel and search for the source language of the table's text — the available languages adapt to whichever OCR engine is currently selected, since not every engine supports every language. Choosing the correct language before extraction noticeably improves character recognition accuracy versus leaving it on the default.

Yes — after extraction, the table opens in an editable grid with low-confidence cells highlighted, so you can correct any misread characters or misaligned columns directly before downloading. Use Reset in the grid to revert to the original OCR output if your edits need a fresh start, or Re-run conversion to try a different engine or language instead.

Start with the Default engine and set Language to match the document, for example French, German or Spanish — table-structure engines paired with the correct language handle most printed invoices well. If columns still come out misaligned or garbled, switch to Engine 1 or Engine 2 with the same language selected, then use the editable grid to hand-correct any remaining low-confidence cells before exporting.

Yes — the verify grid that opens after extraction has an Add row button below the table and a small trash-can icon on the left of every row, so you can insert a line the OCR skipped or remove one created by a smudge, ruled line or stray mark. These row-level edits are baked into the CSV you download alongside any text corrections you make, and hitting Reset discards them together with everything else to restore the original extraction.