LOCAL IMAGE TO TEXT

Image to Text / OCR

Extract printed text in 16 languages from a photo or screenshot, then edit and copy it. Your image stays in your browser.

1. Prepare your image

Choose or drop one PNG, JPG or WebP. Up to 10 MB, 16 megapixels and 8,000 pixels per side. Use a sharp, upright image with clear printed text.

No image selected.

Recognition settings

For mixed-language images, select a second language. Only selected models download. Two languages use more memory and can take longer. Language detection is manual.

Rotate, crop & contrast

Crop coordinates refer to the rotated image. Leave width and height blank to use the remaining image.

Choose an image to begin.

The first run downloads the OCR engine and selected language files from this website. Slow devices can take longer. Large images are reduced to 4 megapixels / 3,000 px per side before recognition.

2. Review the text

OCR can misread names, numbers and punctuation. Check the text against your image before using it.

0 characters

Designed for printed text. Handwriting, complex tables and exact document formatting are not supported. For PDFs, first use PDF to Images and open a page image here.

Using Image to Text / OCR

Turn printed text in a photo or screenshot into editable text. Recognition runs in your browser with locally hosted models for 16 languages.

  1. Choose or drop one PNG, JPG or WebP up to 10 MB, 16 megapixels and 8,000 pixels per side. You can also load the printed-text example.
  2. Choose the main language that matches your image, and optionally a second language for mixed text. Use automatic layout for a page, one block for a paragraph, one line for a label or scattered text for separate labels. Rotate the image upright and crop unnecessary areas if needed.
  3. Extract text and wait for recognition. Compare the result with the preview, correct any mistakes, then copy the text or download a UTF-8 TXT file. Cancel stops processing.

Example

Source: a clear screenshot containing Order number: 12345.
Choose English and One line if the image contains just that label.
Expected text: Order number: 12345. Check the digits before using the result.

Questions & answers

Are my images uploaded?

No. Images and extracted text stay in browser memory. The engine and selected language files download from this website. This tool does not upload or save your image or text. Normal browser caching may retain engine assets; language data is not written to IndexedDB by the OCR engine.

Which languages are supported?

English, Simplified Chinese, Traditional Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Turkish, Korean, Italian, Ukrainian and Hebrew. Combine any two for mixed text. Choose only the languages you need; two models take more memory and time. Languages are selected manually, not detected automatically.

Can it read handwriting or restore a table?

This tool is designed for printed text. Handwriting and complex table reconstruction are not supported. Output is plain text and does not preserve exact fonts, columns or document formatting.

How can I improve the result?

Use a sharp image, straighten the text, choose the correct language and crop unrelated areas. Try grayscale, high contrast or invert for light text on a dark background. Check the prepared preview: strong contrast can erase faint characters.

What does engine confidence mean?

It is an internal OCR estimate, not a measured accuracy score or a guarantee. Even high-confidence output can contain mistakes. Always check important names and numbers.

Why is my image reduced?

Recognition uses at most 4 megapixels and 3,000 pixels per side to limit memory and processing time. Small text in a large page can become unreadable after reduction; crop one section and run it separately.

Can I use a PDF or animated image?

PDFs must first be converted into page images using PDF to Images. PNG and WebP animations are frozen to one still frame; use a static screenshot for predictable results.

Why did recognition stop?

You can cancel at any time. A two-minute timeout stops stalled or very slow recognition. Try a smaller crop, one language or a faster device. Modern browsers with WebAssembly and dedicated worker support are required.

Guides for this tool

Examples and checks to help with your next step.

KEEP YOUR WORKFLOW MOVING

A useful next step for your task.

Browse Images & Design

Help improve this tool

Report a problem or suggest an improvement

Describe the issue without pasting private tool input. Feedback goes to our admin inbox.

Find another tool · Read practical guides