Image & PDF OCR
Turn printed English and Hebrew text into editable text with Tesseract.js. Upload an image or select PDF pages.
זיהוי טקסט באנגלית ובעברית מתוך תמונה או PDF, ישירות בדפדפן.
Your files stay in this browser. OCR runs locally. The tool downloads open-source scripts, WebAssembly and language models from jsDelivr; your document is not sent to an OCR service.
PNG, JPEG, WebP, BMP or PDF. One file, up to 25 MB. Images up to 20 megapixels and 16,384 pixels per side. PDFs: up to 100 pages; OCR up to 20 selected pages per run.
Choose a file to begin.
Recognized text
How to use this tool
Select a clear image or PDF. Choose English, Hebrew, or both. For PDFs, enter page numbers or ranges, then click Recognize text. Every selected PDF page is rendered to an image and OCR runs on that image, even when the PDF already contains selectable text. Results appear page by page. Copy or download your edited combined text.
Use sharp, upright scans with good lighting and clear printed letters. Hebrew is supported through the heb language model, not just translated buttons. Mixed-language results preserve Unicode and line breaks, but reading order, punctuation, tables, handwriting and complex layouts can need correction. Check the result against the original. This does not create a searchable PDF.
Privacy, limits & troubleshooting
Files and recognized text are held in browser memory until cleared or the page closes. No analytics are added by this tool; the surrounding site may have its own analytics. jsDelivr receives ordinary asset download requests; downloaded language models may be cached in your browser. No account or API key is needed. Initial downloads can take time and require internet access.
Use a recent browser with Web Workers, WebAssembly and PDF.js support. Private browsing, blocked CDNs or restrictive browser policies can prevent loading. If a download fails, check your connection and retry. Cancel stops the OCR workers and keeps only already completed page results, marked partial. File or language changes clear old text. No camera permissions are requested.
Maximum file size: 25 MB. Images: 20 megapixels. PDFs: 100 pages total, 20 selected pages per run. PDF pages render at up to 2x scale, capped at 8 megapixels and 8,192 pixels per side; unusually large pages are scaled down, which can reduce OCR accuracy. Encrypted/password-protected PDFs are not supported. Unlock a copy with software you trust before using it here.