Finance & admin

Image table to Excel converter

Get image table entities: visual headers, rows, columns, numeric cells, dates, totals, footnotes, and source-image context.

Before you start

What to expect

Files are retained only for the product retention window and deleted after 24 hours unless your account policy says otherwise.
Uploads are not used to train models; AI Deep Read reads the page only to produce your requested output.
AI Deep Read costs 1 credit per rendered page and still needs human review for financial records.
Image tables are reconstructed from pixels; always review identifiers and totals because OCR confidence is not a guarantee.

An XLSX reconstruction of the table visible in the image, with one rectangular grid per detected table.
Numbers typed as numbers where recognizable, including decimal values, percentages, and negative signs.
Dates normalized when the image makes the format clear, with ambiguous dates left for review.
Headers and row labels preserved so 0/O, 1/l, and wrapped-cell OCR errors are easier to spot.

Input: PNG screenshot of a sales dashboard table, 18 rows, 9 columns, two wrapped product names, percentage columns, and a cropped chart beside the table.
Output: One XLSX sheet with 18 rows and 9 columns; Revenue and Margin typed as numbers, two wrapped product names preserved, and one Product ID needed correction from O to 0.

Cell alignment lost in a screenshot: AI Deep Read infers rows and columns visually, but merged cells and wrapped labels may need manual adjustment in Excel.
OCR confuses similar characters: Review SKU, account, and ID columns for 0/O, 1/l, 5/S, and currency-symbol mistakes before upload to another system.
Photo perspective, glare, or crop: The workflow reads the image as provided. Retake or crop the photo if a row edge or column header is missing.
Multiple unrelated tables in one image: Separate tables may become separate sheets or blocks. Do not assume one screenshot always maps to one clean sheet.

Command line

ocrmypdf

bash

python -c 'from PIL import Image; Image.open("table.jpg").save("table.pdf")' && ocrmypdf table.pdf table-ocr.pdf

Creates OCR text, not a structured workbook; use it as a first step when you need local processing.

pdftotext -layout

bash

pdftotext -layout table-ocr.pdf table.txt

Text layout can help inspect the OCR result, but manual parsing is still required.

Can it convert a screenshot from a web app?

Yes, if the table is readable in the screenshot. Export CSV from the app directly when available; it will be more reliable than OCR.

Does it preserve colors and conditional formatting?

No. The output is a data table. Colors, charts, and dashboard styling are not reproduced as Excel formatting.

What image resolution works best?

Use the sharpest original screenshot or photo. Blurry resized images increase OCR mistakes in digits and small headers.

Can I convert handwritten tables?

Simple handwritten tables may work, but handwriting is less reliable than printed text. Review every cell before using the spreadsheet.