imissfiles55 tools
  1. Home
  2. OCR tools
  3. PNG to JSON

PNG to JSON

Recognise the text in a PNG and export it as JSON, with word positions and confidence.

PNG to JSON

Never uploaded

Your file is processed in this tab and never sent anywhere.

Ads are Google. Files are not.

How to convert PNG to JSON

  1. Add a PNG. A screenshot, an exported graphic, or a lossless scan.
  2. Pick a shape. Plain text, words with coordinates, or inferred rows and columns.
  3. Download the .json. Standard JSON, ready to parse.

About PNG to JSON

PNG is the input format to prefer when you have a choice. It is lossless, so the characters are exactly the pixels that were drawn, with none of the edge softening JPEG introduces. Recognition is fundamentally shape matching, so cleaner shapes mean higher confidence scores and fewer values you have to send for review.

The typical source is something machine-rendered: a screenshot of a dashboard that only exports as an image, a table in a slide, a terminal capture, a UI you are scraping for a test fixture. Because the original was drawn rather than photographed, the text has no perspective distortion and the alignment is genuinely consistent, which is exactly the condition under which position-based table detection works properly.

The output carries the parts a plain text export loses. Every recognised word comes with a bounding box and a confidence score, so you can locate a value by where it sits rather than by matching a label that might change, and you can flag anything the recogniser was unsure about instead of importing it silently. The table shape returns inferred rows and columns as arrays for when the source really is a grid.

One failure mode is specific to screenshots and worth naming: scale. A capture taken at a small size, or one that was resized down before it reached you, may not have enough pixels per character to recognise well. Capture at full resolution where you control it. Enlarging a small screenshot first does not help, because upscaling invents no detail — it just hands the recogniser a blurrier version of the same problem.

PNG to JSON FAQ

Does PNG give better OCR results than JPG?

Generally yes. PNG is lossless, so letterforms keep the hard edges recognition depends on, and confidence scores tend to be higher.

What is in the JSON?

The recognised text with an overall confidence score, or every word with its bounding box and confidence, or inferred table rows — whichever shape you select.

Can I use this in a script?

Yes, that is the point of the format. It is standard JSON that parses anywhere with no special handling.

Why is my small screenshot recognised badly?

Too few pixels per character. Recapture at full size. Upscaling the existing image will not recover detail that was never there.

Is the image uploaded to be recognised?

No. Tesseract runs as WebAssembly in this browser tab and nothing is transmitted.

How is this different from PNG to CSV?

Same recognition, different output. CSV gives you a flat grid. JSON keeps positions and confidence, which is what you want when code is consuming the result.

Related tools