PandaFileKit

Image to text (OCR)

Read the text in a photo, scan or screenshot and copy it out.

JPG · PNG · WEBP → TXT

Sent to server: 0 B

Requests made by this tool while processing · how to verify
Requests that could carry data
0
Engine downloads (code, models)
0
Analytics pings (no file data)
0

Counted with the browser's Performance API during the last job. It sees every request the page makes, but not what is inside them, so anything that is not a static download is counted above.

To check yourself: open DevTools → Network, clear the list, run the tool, and inspect each request. None will contain your file. You can also switch to offline after the page loads; tools whose engine is already downloaded keep working.

Drop images here

or paste, or choose files

JPG · PNG · WEBP → TXT

Choose one or more images of printed text: scans, screenshots, photos of documents.

Copy text from photos, screenshots and scans (OCR)

Image to Text uses optical character recognition to read printed text in an image, so you don't have to retype a notice, a scanned letter or a screenshot. It is also the fix for scanned PDFs where PDF to Text returns nothing.

  • Copying the text of a photographed notice or circular.
  • Getting editable text from a scanned letter or form.
  • Pulling text out of a screenshot that doesn't allow copying.
  • Reading scanned PDF pages after converting them with PDF to JPG.

Pick the language of the text

Recognition works best when it knows the script. English is the default; choose Nepali (or English + Nepali for mixed documents) for Devanagari text. The language data is downloaded to your browser the first time you use it.

  • Photograph documents straight on, in even light, with the page filling the frame.
  • Crop to the text area first; borders and backgrounds cause stray characters.
  • Low-resolution screenshots improve if you enlarge them 2× with Image Resizer first.
  • Several images can be read in one run, and each image's text is kept separate in the output.

Proofread before you rely on it

OCR is good with clean printed text and weak with handwriting, stylised fonts, low contrast, curved pages and small text. Tables come out as lines of text without columns. Common mistakes are 0/O, 1/l/I and missing diacritics or matras. Photos taken at an angle or of curved pages (the middle of a thick book) come out worst; flatten the page and shoot from directly above.

  • Double-check numbers, names, dates and amounts against the image.
  • Check line breaks where columns may have been merged.

Image to text (OCR): questions

Can it read Nepali or Hindi text?

Choose Nepali for Devanagari text, or English + Nepali for mixed documents. The language data downloads the first time you use it.

How do I get text from a scanned PDF?

Convert the PDF pages to images with PDF to JPG, then add those images here.

Can it read handwriting?

Only poorly. It is designed for printed text; neat block capitals work better than joined writing.

How can I improve accuracy?

Use a sharp, straight-on photo in even light, crop to the text, and enlarge small screenshots 2× before reading them. Always proofread numbers and names.

Are my images uploaded for OCR?

No. Text recognition runs in your browser, so scanned letters and documents stay on your device.