Skip to content
Utilboxes

Image to Text (OCR)

To copy text out of an image, add a screenshot, photo or scan, choose its language and press Extract text, then copy or download the result. Recognition uses the Tesseract OCR engine inside your browser, so the image is never uploaded. Clear printed text is usually over 95% accurate; handwriting and angled photos are much less reliable.

Files never leave your device

Updated

How to use Image to Text (OCR)

  1. 1Add an image containing text.
  2. 2Choose the language of the text, which improves accuracy considerably.
  3. 3Press Extract text and watch the progress — the first run downloads the language data.
  4. 4Review the result, then copy it or download it as a .txt file.

How it works

This uses Tesseract, the long-established open-source OCR engine, compiled to WebAssembly and run inside a Web Worker in your browser. The engine finds text regions, segments them into lines and characters, and matches character shapes against a trained model for the language you selected.

Accuracy depends almost entirely on the input. A flat, well-lit screenshot of printed text is usually near-perfect. A photograph taken at an angle, in poor light, or of handwriting will be much less reliable — Tesseract is trained on printed type, not handwriting.

Choosing the right language matters more than people expect, because the model uses a dictionary to resolve ambiguous shapes. Running English text through the German model will produce noticeably worse output.

The language data is a few megabytes and is downloaded once, then cached for later runs. Recognition itself happens entirely on your device — the image is never sent anywhere.

Tips and common mistakes

  • Make the letters big enough: capitals around 20–30 pixels tall or more. Zoom in before taking a screenshot rather than enlarging a small one afterwards.
  • Crop each column of a two-column page into its own image; OCR reads straight across the gap otherwise.
  • Proofread look-alike characters — 0 and O, 1 and l, 5 and S — and every number you will rely on.
  • Paste the result into Text Cleaner and join broken lines to turn the line-by-line output back into paragraphs.

Frequently asked questions

How accurate is the text recognition?
On clear printed text — screenshots, PDFs rendered as images, flat scans — accuracy is typically well above 95%. Angled photographs, low light, unusual fonts and handwriting all reduce it sharply.
Can it read handwriting?
Not reliably. The engine is trained on printed type. Neat block capitals sometimes work; ordinary cursive rarely does.
Which languages are supported?
A range including English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Chinese, Japanese, Korean, Arabic and Hindi. Each downloads its own trained model on first use.
Is my image uploaded to a server?
No. The OCR engine runs inside your browser tab. Only the language model is downloaded, and that is the same public file for everyone.
How can I get better results?
Use the highest-resolution image you have, make sure the text is horizontal and in focus, crop away irrelevant background, and pick the correct language.
Can it read the text in a scanned PDF?
Turn the PDF's pages into images with PDF to PNG first, then run each page image through this tool.
Why are the cells of a table jumbled?
OCR returns text line by line and does not know where table cells begin and end. You get the words and figures, but the grid has to be rebuilt by hand.
How long does recognition take?
Usually a few seconds for a screenshot on a recent device. The first run also downloads the language data, and large or detailed images take longer.
Share this toolShare on XShare on FacebookShare on LinkedInShare on RedditShare by email

Guides

Related tools

PDF to TextExtract the plain text from a PDF, page by page.PDF ToolsWord CounterLive word, character, sentence and reading-time counts.Text ToolsImage to PDFCombine any mix of images into one PDF document.Text CleanerFix smart quotes, stray spaces and broken line breaks.Text Tools