Imagen → texto (OCR)
Extrae texto impreso
Sobre esta herramienta
Extract text from images using on-device OCR powered by Tesseract.js. Works on screenshots, photos of documents, signs, scanned receipts, hand-drawn whiteboards, and pages from books - in over 100 languages.
Because the OCR engine runs entirely in your browser via WebAssembly, you can scan confidential documents, medical paperwork, contracts and ID cards without anything leaving your machine. Copy the text to clipboard or download as a .txt file.
OCR runs locally via Tesseract compiled to WebAssembly. The image is binarized, segmented into text blocks and recognized character-by-character against a trained language model. Accuracy depends heavily on image quality: 300 DPI scans of printed text usually exceed 98% accuracy, while photos of handwriting or low-contrast signs can drop into the 60-80% range.
Características
- Tesseract.js OCR running locally in your browser
- 100+ language packs (download on demand)
- Reads JPG, PNG, WebP, BMP and most image formats
- Copy text to clipboard with one click
- Download recognized text as .txt
- Confidence scores per word (advanced view)
- Browser-only - images never uploaded
Cómo usarla
- Drop in an image with text.
- Pick the document language (or 'eng' for English).
- Click Recognize.
- Copy the extracted text or download as .txt.
Todo sucede dentro de tu navegador usando JavaScript y WebAssembly. Tus archivos nunca se suben a un servidor, nunca se almacenan y nosotros nunca los vemos.
Preguntas frecuentes
OCR accuracy depends on contrast, resolution, font and skew. Higher-resolution images, cleaner crops, and good lighting all help. Handwritten text is the hardest - printed text usually scans well.
Over 100 - including English, Danish, German, French, Spanish, Italian, Swedish, Norwegian, Polish, Russian, Chinese, Japanese, Arabic and more. Pick the right one for best results.
OCR runs per image. For PDFs, convert pages to images first with the PDF → Images tool, then OCR each page.
Lens is more accurate on hard images (handwriting, skewed text) but uploads your image to Google. Privvert is local, free, and accurate enough for most printed text.
It works on both, but scans win. For phone photos, hold the camera parallel to the page (no perspective skew), make sure lighting is even, and crop to just the text area before running OCR.
Cloud OCR services see every word in every receipt, ID card, lease or contract you upload. Running OCR in the browser means those documents never leave your device - the same reason the redaction article argues for offline tools when handling sensitive paper.