PDF to XLSX uses AI Deep Read to read page images, emit Markdown tables, and then let our table engine write an Excel workbook.
From real runs
| Element | Result | What happens |
|---|---|---|
| PDF prose | Changed | The title and explanatory paragraph became single-cell rows above the extracted table in sheet1.xml. |
| Table headers | Kept | invoice, customer, qty, amount, and note were written across columns A through E. |
| Text cells | Kept | Invoice IDs and customer names, including Café Norte, were preserved as string cells. |
| Numbers | Kept | Quantities and amounts were written as numeric cells, including 2, 1234.5, 4, and 88. |
| Accounting negatives | Changed | The PDF value (19.95) was normalized to numeric -19.95 in the generated XLSX worksheet. |
| Sheet naming | Changed | The workbook used Sheet1 in the API result rather than a document-derived name such as Vendor credit memo. |
| Original layout | Lost | PDF page geometry, column widths, fonts, and page breaks were not represented in the Excel output. |
The title and explanatory paragraph became single-cell rows above the extracted table in sheet1.xml.
invoice, customer, qty, amount, and note were written across columns A through E.
Invoice IDs and customer names, including Café Norte, were preserved as string cells.
Quantities and amounts were written as numeric cells, including 2, 1234.5, 4, and 88.
The PDF value (19.95) was normalized to numeric -19.95 in the generated XLSX worksheet.
The workbook used Sheet1 in the API result rather than a document-derived name such as Vendor credit memo.
PDF page geometry, column widths, fonts, and page breaks were not represented in the Excel output.
We converted our PDF sample (91 KB) with AI Deep Read under AI Deep Read on 2026-09-11. It took 17 s and produced a 17 KB Excel XLSX file, 81% smaller than the input. The engine raised 1 warning, listed below.
Input · sample.pdf · 91 KB
PDF 1.4, 1 page, no text layer, 1 image object
Output · sample.xlsx · 17 KB
Workbook with 1 sheet (Extracted); first sheet 13 rows, 24 cells, 0 formulas, header filter, fitted column widths
A scanned PDF will not have selectable table text and usually has one image-heavy object per page. AI Deep Read spends vision credits by page, reads the page image, and then asks our table engine to materialize detected pipe tables. Review totals, merged labels, and accounting negatives against the original.
pdftotext plus spreadsheet cleanup
pdftotext -layout table.pdf table.txtThis command is useful for checking a digital text layer, but it does not produce typed Excel cells or OCR scanned pages.
The HTTP API and the MCP server run the exact engine this page uses (AI Deep Read); the response carries the same warnings the web converter shows. Check support first with GET /api/v1/pairs?from=pdf&to=xlsx.
curl -X POST https://filetypeconverters.com/api/v1/convert \
-H "Authorization: Bearer $FTC_API_KEY" \
-F "file=@input.pdf" -F "from=pdf" -F "to=xlsx" \
-F "deepRead=true" \
--output output.xlsx/convert/pdf-to-csv when a plain import file is enough and workbook packaging is unnecessary./convert/pdf-to-json when you need paragraphs and detected table Markdown for a downstream extraction pipeline.Deep Read returns page text plus tables. In the tested file, the title and explanatory sentence became rows above the table in Excel.
In our run, quantities and amounts were numeric cells, and the accounting negative was a numeric -19.95. Review IDs separately because invoice numbers stayed strings.
No. The output is a spreadsheet interpretation of detected text and tables, not a fixed-layout copy of the PDF page.