PDF tables to Excel or CSV made simpler, in your browser

A supplier list, expense statement, or inventory report can be easy to read in a PDF and difficult to reuse in a spreadsheet. Copying an entire page often mixes headings with data or places several columns in one cell. This tool gives you control over a smaller task: select the table you need, choose its page range, inspect the extracted rows, and save those rows as an Excel workbook or CSV file.

The output contains reviewed cell text, not a replica of the document's visual formatting. It does not reconstruct formulas, merged cells, charts, fonts, or colored table styles. That makes the result useful for moving a straightforward text table into your own workbook. Before using the values in a calculation, compare the extracted data with the source, especially account identifiers, decimal points, negative amounts, and totals.

Choose one PDF and use the preview page control to find the table. Drag from one corner of the desired area to the opposite corner. You can also enter left, top, width, and height as percentages of the visible page. These numeric controls work with a keyboard and let you make a more precise selection when a table is close to a heading, footnote, or neighboring paragraph.

Keep all the wanted cells inside the region, including the first column and the bottom row. Leave unrelated captions outside where possible. Selection follows the page as you see it, including its displayed rotation and visible crop boundary. A selected text block is included by its center position, so a box that cuts through letters may exclude more than you expect. Compare the first extraction with the source and widen the region if data is missing.

A PDF often records pieces of positioned text rather than a ready-made table. Automatic column detection looks for repeated starting positions across rows. This can work well for a simple list, but right-aligned amounts, unusually long descriptions, or inconsistent spacing can produce an imperfect result. The editable preview is therefore part of the task, rather than an optional final glance.

If automatic columns do not fit, enter your own column boundaries. For example, 35,70 makes three columns within the selected region. These percentages are measured from the region's left edge, not the whole page. Boundaries must increase from left to right and stay between zero and one hundred. The summary also reports text blocks that cross a boundary. Move the boundary into a gap between cells and extract again when a description belongs to the wrong column.

Yes. Enter a page range such as 1,3-5 to extract the same region from several physical pages. The tool processes those pages in document order and appends their rows into one sheet. Use this when a long table continues in roughly the same position, with the same column arrangement, on each page. A cover counts as a page, regardless of the printed number in the footer.

Repeated headers are kept as ordinary rows so that no heading is silently discarded. Delete the repeated rows in the preview if you do not need them. If different pages place the table in different areas, extract those sections separately rather than forcing one rectangle across incompatible layouts. Changing the preview page resets the proposed extraction range to that page; re-enter a wider range when you deliberately want several pages together.

Characters belonging to the same visual row can have slightly different baseline positions. Row alignment tolerance controls how close those positions need to be before their text is grouped together. The default is four points, and you can choose from one to twenty. Increasing it can join items that were incorrectly split into two rows; decreasing it can separate nearby rows that were combined.

Large tolerance values can merge genuine neighboring rows. Begin with the default and adjust only when the preview shows a specific alignment problem. A multi-line description is not automatically recognized as one logical record: its second line may appear as another row. Move or combine that text in the editable cells before exporting. The tool does not infer which transaction a wrapped sentence belongs to from its meaning.

This extractor needs an existing horizontal text layer. A page that is only a photograph has pixels, but no positioned characters to place in spreadsheet cells. If no suitable text is found, the tool reports that limitation rather than producing an empty workbook. You can use the OCR tool first to create a searchable copy, then return here and inspect what the recognized text positions support.

OCR does not reliably rebuild table structure. Some recognized lines become one wide text block, and recognition can confuse similar digits or lose punctuation. A scan may therefore need substantial manual correction, or a dedicated table recognition workflow, even after OCR. For amounts that matter, compare each row with the original image. This page does not promise automatic reconstruction of scanned financial statements, handwriting, or complex nested tables.

The Excel export stores every cell as text. An identifier such as 001234 therefore keeps its leading zeroes, and a long reference number is not deliberately converted into a floating-point value. Dates, percentages, currencies, and negative amounts are also stored as their displayed text. This is a useful starting point when accurate source representation matters more than automatic calculation.

After opening the workbook, convert only the columns that should be numeric. Check decimal and thousands separators for your spreadsheet's locale before doing that conversion. CSV import depends on the application opening the file, so choose a text column type for identifiers when using an import dialog. Simply double-clicking a CSV can let a spreadsheet apply its own number or date interpretation, which this website cannot control.

Choose XLSX when you want a workbook with one worksheet containing your reviewed table. It opens directly in compatible spreadsheet applications and explicitly stores cells as text. Choose CSV when another system needs a plain, comma-separated data file. The CSV includes a UTF-8 marker and quotes cell values, including commas, quotation marks, and line breaks, so those characters are represented consistently.

Neither format includes source-page styling or formulas. For CSV, content that resembles a spreadsheet formula receives a leading apostrophe to help prevent accidental execution when opened in a spreadsheet. This can also affect a negative number represented as text. Review those values during import and convert intended numeric columns deliberately. XLSX uses text cells without formula elements, so an expression such as =2+2 remains literal content.

Click a cell in the preview to edit its text. You can delete an unwanted row or add a missing row, then fill its cells. Check the column count before making many corrections: a better region or boundary setting may resolve the same problem across every row. Editing a cell invalidates the previous download, so export again when you have finished reviewing the changes.

For a report with totals, compare the number of records and a few independently checked sums after importing into your spreadsheet. Be careful with duplicate headers, blank-looking cells, and footnotes that resemble records. The extractor does not validate business rules or certify that the copied values are complete. Retain the source alongside your workbook so another person can trace a doubtful value back to the document.

Choose one unencrypted PDF up to 30 MB and one hundred pages. A table export is limited to one thousand rows, twenty columns, and five thousand cells, with up to ten thousand characters per cell. If the selection exceeds these limits, use a smaller page range or table region. Very large pages may also exceed the browser's safe preview pixel limit.

Processing takes place in your browser; the selected PDF is not uploaded to a conversion service. You can clear the file, cancel extraction, or choose another source. Horizontal text tables are the intended input. Vertical writing, decorative text, and page layouts that do not expose usable character positions may need a different workflow. Open the final workbook or CSV and check its content before sharing it or using it for analysis.

Choose a tool for the next step and review each output.

• Add searchable text to a scan first
• Extract ordinary paragraphs