Extract Table

Location on ribbon: Tab DVD Addin → Group File and Print → Menu PDF / OCR

Command ID: btnExtractTable

License required

This command needs an active DVDAddin account (Standard or PRO). During the trial period it is fully available.

What it does

This command recognises a table inside a PDF or image and rebuilds its row/column structure in Excel. Unlike plain OCR, which pours text down one column, the model is asked to return a single HTML table in which every row has the same column count and merged cells are reproduced with rowspan/colspan — so a bill of quantities or a multi-level lab-result header comes back in the right shape.

Before anything is pasted, a preview window lets you check and correct each cell. Clean numbers are converted to real numbers so Excel can total them immediately instead of treating them as text.

How to use

1. Select the target cell — it becomes the top-left corner of the pasted table.
2. Go to tab DVD Addin → group File and Print → menu PDF / OCR → click Extract Table.
3. A file picker opens, filtered to .png, .jpg, .jpeg, .webp and .pdf. Choose the file and click Open.
4. If the PDF has more than one page, the add-in asks which pages hold the table. Pick only the pages you need — it is faster and cheaper than sending the whole document.
5. The status bar shows that extraction is running.
6. The preview window opens with the recognised grid. Fix any misread cell in place. When the source had merged cells you also get an option to merge cells the same way, plus an option to add borders to the pasted range.
7. Click OK. The whole grid is written into the sheet in one pass, starting at the cell you selected.

Example

The consultant sends rebar tensile test results as a PDF. The table has two header tiers (Vietnamese above, English below) and a "Sample" column merged across three rows per sample set.

You select A6 on the TestResults sheet, run Extract Table, pick the PDF and choose page 2. The preview shows a 14 × 8 grid: header cells such as "STT / No", "Đường kính / Diameter" and "Giới hạn chảy / Yield" each sit in a single cell, and "Φ 12 – welded mesh" spans three rows as in the original. You correct one misread 0,58 to 0.58, tick merge and borders, and click OK — the table lands in A6:H19, with the strength columns as real numbers so SUM/AVERAGE work right away.

Notes

• A valid API key and an internet connection are required; without a key the command stops before doing anything.
• The paste overwrites existing data from the active cell onward and cannot be undone with Ctrl+Z — paste into an empty sheet first when in doubt.
• Merged cells are only reproduced when you do not add or remove rows/columns in the preview; changing the structure shifts the stored merge coordinates, so the add-in skips merging in that case.
• If the model does not return valid HTML, the add-in falls back to reading the result as tab-separated text; if there is still no table it reports that none was found.
• Slanted printouts, broken rule lines, or phone photos with curved pages cause column drift — always reconcile totals against the source before using the data in a payment application.

Related

• OCR file — same AI engine, but returns running text rather than a grid.
• OCR Screen Clip — grab content from a region of the screen instead of a file.
• Read PDF — view, mark up and work with PDF files inside Excel.
• PDF Manager — split, merge or rotate pages before extracting a table.