boxtool.io

OCR — Image to Text

Extract text from images using Tesseract OCR. 12 languages. Runs in your browser — no upload, free.

Drop an image or click to browse

PNG, JPG, WEBP, BMP, GIF, TIFF

All processing happens in your browser — your images never leave your device. First run downloads language data (~4 MB/language), cached afterward.

About OCR — Image to Text

Optical Character Recognition (OCR) is the technology that converts images of text — scanned documents, screenshots, photos of signs, or any image containing readable characters — into actual machine-readable and editable text. It eliminates the need to manually retype text visible in images, saving significant time when working with scanned contracts, receipts, printed notes, or screenshots of content that cannot be selected.

Our OCR tool runs Tesseract — the leading open-source OCR engine, originally developed at HP, sponsored by Google from 2006 to 2018 and maintained by the open-source community today — as a WebAssembly module directly in your browser. This is significant: your images are never uploaded to any server. Tesseract supports 12 languages including English, Portuguese, Spanish, French, German, Italian, and more, with language data downloaded once and cached locally for subsequent uses.

Accuracy depends heavily on image quality. High-contrast, sharply focused images with clean typography produce near-perfect results. Handwritten text, stylized fonts, poor lighting, and heavy compression artifacts reduce accuracy. For best results, use images where text is clearly legible to the human eye — if you can read it easily, Tesseract can usually extract it accurately.

How to Use OCR — Image to Text

  1. 1Drop a PNG, JPG, WebP, BMP, GIF or TIFF image into the upload area, or click it to browse.
  2. 2Select the language of the text in the image from the language dropdown — 12 are available.
  3. 3Click "Extract Text" to start the OCR process — a progress bar reports the recognition percentage, and the first run may take a moment to load language data.
  4. 4The extracted text appears in the output panel and is editable for corrections.
  5. 5Check the confidence score shown next to the output heading to judge how much proofreading it needs.
  6. 6Click "Copy" to copy the extracted text or "Download" to save it as a .txt file.

Use Cases

  • Extract text from a scanned document, invoice, or receipt to make it editable
  • Copy text from a screenshot of a website, PDF, or article that cannot be selected
  • Digitize printed notes, textbook pages, or printed documents for editing or searching
  • Read and capture text from photos of signs, menus, product labels, or printed materials
  • Turn a scanned PDF into text by exporting its pages as images with our PDF to Image tool first, then running OCR on them
  • Extract data from images of tables or forms for import into a spreadsheet
  • Fix an OCR mistake right in the output panel before copying — the extracted text is editable

Tips

  • High-contrast, sharp images produce significantly better OCR accuracy than blurry or low-resolution ones
  • The first recognition run may be slower as the OCR engine (~10 MB) and language data (~4 MB) download and cache locally
  • The confidence score is an average across the page — a low number usually means a blurry or skewed scan
  • The tool reads images only; convert PDF pages to images first with the PDF to Image tab
  • Select the correct language before starting — accuracy drops significantly with the wrong language model
  • Increase image resolution or zoom before taking a screenshot to improve OCR quality
  • Handwritten text has much lower accuracy than printed text — results may need manual correction
  • For multi-column documents, the extracted text follows reading order but may need reformatting

Frequently Asked Questions

PNG, JPG/JPEG, WEBP, BMP, GIF and TIFF. For best results use high-resolution images with clear, high-contrast text.

English, Portuguese, Spanish, French, German, Italian, Dutch, Polish, Russian, Japanese, Chinese (Simplified) and Arabic.

For clean, printed text with good contrast, expect 95–99% accuracy. For handwritten text or low-quality scans, accuracy will be lower.

No. OCR runs entirely in your browser using WebAssembly (Tesseract.js). Your images never leave your device.

Not directly — the tool accepts image files. Convert PDF pages to images first, then run OCR on each image.

On first use, the OCR engine (~10 MB) and language data (~4 MB) are downloaded and cached. Subsequent recognitions start immediately.

Related Tools

Guides & reading

Ad