Image to text
Extract printed text from an image, then correct and copy it.
Files stay on your device. Nothing is uploaded.
One PNG, JPEG, WebP or BMP; up to 10 MB and 8 million pixels.
The OCR engine and selected language download after you start. No image is uploaded.
Review names, numbers and reading order. Handwriting and complex tables may not be recognized correctly.
Related tools
-
AI image upscaler
Enlarge a small picture 2× or 4× with a model running on your device. -
Word counter
Live word, character, sentence and paragraph count as you type. -
Online notepad
Write, autosave and download plain text notes without an account. -
Image and PDF character counter
Count poster and PDF text, review each page, then share the totals.
Copy printed words from a screenshot or scanned image without typing them again. This OCR tool recognizes text locally in your browser and places it in an editable field. Choose the language of the document, not necessarily the language of this page. Clear, upright text with good contrast works best. The result is plain text rather than a recreation of the document layout: check line breaks, numbers and names before using it. Your image stays on your device throughout recognition.
How to use
- Choose one PNG, JPEG, WebP or BMP image up to 10 MB and eight million pixels.
- Select the printed text's language, then start recognition; the page downloads only the selected language model.
- Compare the editable result with the image, fix mistakes and copy it or download a plain TXT file.
What it handles
- Seven language models for English, Ukrainian, Spanish, French, German, Portuguese and Italian.
- Plain editable text, suitable for notes or pasting into a document; no hidden formatting is retained.
- Local recognition with a visible progress status and no file upload or account requirement.
- Clear removes the file and recognized text; results are not automatically saved between visits.
FAQ
Are my files sent to an AI service?
OCR reads the screenshot locally with the selected language model. Your file is processed in this browser, not uploaded to an inference service. Required models and runtime files are downloaded from this site after you start processing. The tool does not save your input or result in local storage. Keep the original file yourself and download any result you want to retain before closing the page.
Does this preserve tables, handwriting or document formatting?
The output is plain text, not an editable copy of the original page layout. Columns and tables can be read in an unexpected order, and handwriting is not a reliable use case. Start with upright printed text, crop unrelated areas and inspect the result. For important numbers, identifiers or quotations, compare every character with the source.
Why does recognition take time or return the wrong words?
The browser first downloads the OCR engine and selected language data, then processes the image on your device. A large image or slower phone can take longer. Choose the actual text language and use a sharp, well-lit image. Tiny letters, compression artifacts, curved pages and mixed languages can reduce recognition quality; no accuracy percentage is guaranteed.
Updated