PNG to CSV

How to convert PNG to CSV

1. Add a PNG. A screenshot of a table, a chart legend, or a scanned grid.
2. Review the detected grid. Rows and columns are inferred from where the text sits.
3. Download the .csv. Opens in any spreadsheet, parses in any language.

About PNG to CSV

This is the best-case input for turning a picture of a table into data, and it is worth knowing why. PNG is lossless, so the letterforms in a screenshot are exactly the pixels the screen drew — no compression softening, no artefacts around the edges. Optical character recognition is a shape-matching problem, and clean shapes recognise better. Expect noticeably higher accuracy here than from a JPEG of the same table.

The common case is a table you can see but cannot copy: a figure inside a dashboard that only exports as an image, a grid in a slide deck, a report someone screenshotted and pasted into a message, a web table behind a login that resists selection. The data is right there and completely inert.

Tesseract recognises the text as WebAssembly in this browser tab, keeping the bounding box of every word. Words are grouped into rows by vertical position and columns are inferred from the x positions where words consistently begin. Screenshots do well at this because the alignment is machine-generated and therefore genuinely consistent — a photographed table has slight perspective that a screenshot never has.

The one thing that hurts a screenshot is scale. A table captured at a small size on a high-density display, or a screenshot that was resized down before you got it, has too few pixels per character for reliable recognition. Capture at full size where you can, and do not enlarge a small screenshot before running it — upscaling invents no detail and simply gives the recogniser blurrier shapes to work with.

PNG to CSV FAQ