Converting an image of a table into a spreadsheet

The three routes

1. A free online image-to-Excel converter

Upload the image, get a file back. Fast, free, and works from any device. The trade-off is unavoidable and usually unstated: the document is uploaded to somebody else's server. For a photo of a conference agenda that is fine. For a bank statement, a payslip or a customer list it is a disclosure, and free services differ enormously in how long they keep uploads and who can reach them. If you use one, read what it says about retention before you drag the file in — and assume anything unstated is unbounded.

2. Excel's own Data from Picture

Microsoft ships a feature that reads a table out of a picture and inserts it into the sheet. It has been available in Excel for Microsoft 365 and in the Excel mobile apps; availability varies by version and platform, so check yours before planning around it. Where it is available it is a good option: it is in the tool you already have, and it is covered by whatever agreement your organization already has with Microsoft.

3. A desktop app with your own AI key

TableDI is a desktop spreadsheet workspace that runs entirely on your machine, and it can import a photograph of a table as rows. Be precise about what "on your machine" covers here: the workspace, the tables and the files are local, but reading an image takes a model, so the image is sent to the AI provider whose API key you entered. It goes from your machine to them. It never passes through anything of ours — TableDI has no server in that path and never sees the key or the image.

That is a meaningfully different position from an anonymous converter site — you chose the provider, you hold the account, and their terms are ones you agreed to — but it is not the same as offline OCR, and we are not going to call it that.

Which one for which document

Bank statement to Excel, specifically

This is the most-searched version of the task and the one with the worst risk profile, so it is worth separating out. Two things to try before any converter:

• Ask the bank for CSV. Almost every bank offers CSV, OFX or QIF export somewhere in the statements screen — often two clicks from where the PDF download is. A structured export is more accurate than any OCR, because nothing is being guessed.
• If the PDF has real text, it is not an OCR problem. Select the text in a PDF viewer; if it highlights, the numbers are already characters and can be copied. Only a scanned page needs a model to read it.

The paid converters that dominate these results exist because bank statement layouts are genuinely hard — multi-line descriptions, running balances, page breaks mid-transaction. If you convert statements every month for many accounts, one of them may well be worth the money. For a handful of pages, the bank's own CSV is better than all of it.

How the import works in TableDI

1. Add your AI provider's API key in Desktop settings (Cmd+, on macOS / Ctrl+, on Windows). Nothing AI-related happens before this and no key is bundled. Setting up a key.
2. Upload file on the home screen, or Import in a table's toolbar, and choose the image.
3. The rows come back as a table, editable like any other — fix what the model misread before you build anything on it. Always check the totals against the document; OCR gets decimals and column alignment wrong often enough that a spot check is not optional.
4. From there it is an ordinary table: formula columns, a pivot, a chart, an export.

Image import is part of the AI plugins, which are on the paid tier. The free tier imports Excel, CSV and JSON with no row cap. The split.