Image to JSON (OCR)
Developers and data teams often need more than plain text: where each line is on the image and how sure the engine was. This tool outputs exactly that as JSON.
- Never uploaded
- Instant, no sign-up
- Works offline
About this tool
Each paragraph includes its text, bounding box and lines, and each line has its own box and confidence — handy for prototypes, form parsing and checking OCR quality, without sending data to an API.
How it works
- 1
Add an image with text (or try the sample).
- 2
Press Extract text — the first run downloads the recognizer once.
- 3
Copy or download the result.
Features
- Bounding boxesPixel coordinates for every paragraph and line.
- Confidence scoresPer-line confidence from 0 to 100.
- No API keysFree, unlimited and private — it runs in your browser.
- Private by designEverything runs on your device. No uploads, no account, no watermark, and EXIF/GPS data is stripped from the result.
Supported formats
Input: JPG, PNG, WebP, BMP, GIF, TIFF and HEIC images of printed text in English. Output: plain text, Markdown, HTML, JSON or CSV.
Frequently asked questions
Are the coordinates in original pixels?
Yes, boxes are mapped back to the original image size.
Can I automate this?
This page is interactive. For batch automation, the open-source tesseract.js library offers the same engine.
Is my image sent to an OCR service?
No. The text recognizer (Tesseract) is downloaded from this site once and runs entirely in your browser. Your image never leaves your device.
Are my images uploaded to a server?
No. Files are decoded, processed and re-encoded by your browser using the canvas API and Web Workers. Nothing is sent to any server, and there is no account or storage. Closing the tab discards everything.
Related tools
Need more than one tool?
The full editor combines cropping, adjustments, filters, text, shapes and layers in one place.