PDF to Excel Converter

The tool reads each text run's real position on the page from the PDF's own content stream, then groups runs into rows by shared vertical position and into columns by shared horizontal start position — the same technique used by well-known open-source PDF table tools. It works well on clean, evenly-spaced tabular data (invoices, exports from spreadsheets, bank statements). It does not detect tables via drawn ruling lines, so very unusual layouts may need manual cleanup afterward.

When you choose "Preserve formatting," the tool detects bold and italic text directly from the PDF's own font information and recreates it in the spreadsheet, along with light cell borders. This depends on a formatting library loading successfully — if it doesn't (rare, but possible on an unreliable connection), the tool says so plainly in the conversion report and exports the data unstyled rather than silently claiming formatting that isn't actually there. Choose "Data only" if you don't need styling and want the fastest, simplest export.

Yes — the tool looks for real gaps in the vertical spacing between rows to tell genuinely separate tables apart, rather than treating everything on a page as one table. Two unrelated tables on the same page are kept distinct instead of being merged into one misaligned grid.

Yes — pages with no selectable text are automatically run through on-device OCR (the same engine as the OCR Image to Text tool), then the same row/column reconstruction is applied to the recognized words. OCR accuracy depends on scan quality; low-resolution scans may extract less reliably.

Dates are kept as readable text rather than converted to Excel's internal date format. Date text is genuinely ambiguous (03/04/2026 could mean different days in different countries) — guessing wrong would silently corrupt a date in your spreadsheet, so the tool preserves the text exactly as printed instead.

Yes — if a PDF needs a password to open, the tool will ask for it before processing. The password is only used locally to decrypt the file in your browser; it is never sent anywhere.

It approximates section-header-style merged rows (a single wide label spanning what would otherwise be several columns) as a real merged cell in the output. This is a best-effort heuristic, not exact PDF structural data, since PDFs don't expose table cell-merge information directly.

Excel (.xlsx) with formatting preserved, Excel (.xlsx) with data only for a simpler file, or plain CSV. All three come from the same table-reconstruction pass — the choice only affects how the result is packaged, not how tables are detected.

It's rule-based structural analysis, not a trained AI model — the tool measures real signals (how short the page is, whether a large title dominates, whether column positions genuinely repeat across many rows) and reports a plain High/Medium/Low confidence with the actual reasons, rather than an invented precise percentage. It reliably tells short documents, table-structured documents, and long-form prose apart; it doesn't attempt to guess a specific document genre (invoice vs. price list vs. bank statement all share the same underlying grid structure).

Smart Auto detects each page's structure and reconstructs it accordingly — recommended for most PDFs. Preserve Visual Layout always groups content into readable blocks instead of a grid, which suits business cards, resumes and other non-tabular documents. Editable Spreadsheet always uses the table-grid path, best when you plan to edit or run formulas on the result. Tables Only extracts exclusively what was classified as tabular content, skipping surrounding text. Text Only is a simple reading-order dump with no grid at all.

Not currently — the underlying spreadsheet library the tool uses doesn't support writing images into cells. Images in the source PDF are skipped rather than silently dropped without mention; this is a real, known limitation, not an oversight.

Yes — completely free, with no usage limits, no watermarks and no premium wall. Every tool on Tarumak Studio is free to use.