Table Image to CSV
Turn a photo or screenshot of a table into CSV: on-device OCR reads the words and their positions, rebuilds the grid as a layout estimate you can edit cell by cell, and exports CSV. The image never leaves your device.
The grid below is a layout estimate rebuilt from word positions. Merged cells, multi-line cells, and skewed or rotated scans will not reconstruct perfectly — review and edit the cells before downloading.
How to use
- Choose an image of a table — a photo or a screenshot.
- Select the recognition language: English, Japanese, or both.
- Press "Extract table". Words and their positions are read with OCR, and the grid is rebuilt from the word layout.
- Check the cells, fix any mistakes, and download the result as CSV.
Notes
Extraction runs with Tesseract.js inside a Web Worker in your browser; the image is never uploaded. The first run downloads the OCR engine and language data from the Tesseract CDN (jsDelivr), cached afterwards. The grid is a layout estimate: rows are grouped by vertical overlap and columns by horizontal word clusters, so merged cells, multi-line cells, and skewed or rotated scans will not reconstruct perfectly. Cells whose text starts with =, +, -, or @ are prefixed with an apostrophe in the CSV so spreadsheet apps do not treat them as formulas.
FAQ
- How accurate is the table reconstruction?
- It depends on the image. Clean, straight-on screenshots of simple grids work best. Ruled lines are ignored — only word positions are used — so borderless tables work too, but wide gaps inside one cell can split it into two columns.
- Is my image uploaded?
- No. Recognition runs entirely in your browser. Only the OCR engine and language data are downloaded, on the first run.