Table OCR - Image to Excel Online

AI rebuilds the table structure, OCR fills in the text - export to Excel, all local

Retyping a table from a report screenshot, a scanned invoice, or a book page is tedious. This free tool combines two local AI models: SLANet-plus reconstructs the table structure - rows, columns, and merged cells (colspan/rowspan) - while PP-OCR reads the text in each cell and drops it into the right place.

After recognition, detected cell boxes are outlined on the source image and the rebuilt table is previewed side by side. Export three ways: copy as tab-separated text (paste straight into Excel or Google Sheets), copy the HTML table markup, or download an .xlsx file with merged cells preserved.

The key difference from cloud table-OCR services: the models run inside your browser, so the image never leaves your computer - financial statements, invoices, and bills are safe to process. Free, no sign-up, no usage limits.

How to use

Click "Table recognition" to open the tool
Upload a table screenshot, photo, or scan (PNG/JPG/PDF)
Click "Recognize table" and let the local AI rebuild the structure and text
Review the preview, then copy into Excel or download the .xlsx file

FAQ

Are merged cells supported?

Yes. The structure model predicts colspan/rowspan, and both the HTML preview and the exported xlsx keep merged cells intact.

What kind of images work best?

Printed tables with clear borders and upright text - like system screenshots or scanned reports. Borderless tables, angled phone photos, and handwriting reduce structure accuracy, so review the result.

Is my image uploaded to a server?

No. Both models run locally in your browser via WebAssembly/WebGPU. The image and the results never leave your device.

What if the result is not perfect?

Copy it as tab-separated text into Excel and fix a few cells there - when the structure is mostly right, that is far faster than retyping the whole table. Larger screenshots with complete borders also noticeably improve accuracy.