Text PDF or scanned PDF?
If you can select the words in a PDF with the mouse, it probably contains a text layer; direct extraction is usually faster and more accurate. A scanned page that behaves like one large image needs OCR.
From photos, screenshots, and scanned PDFs, Image to Textand PDF OCR run directly on your device. PDF pages with selectable text are read from the text layer instead of using OCR.
JPG, PNG, WebP, BMP, PDF · Original files are not uploaded to the MISCLAND server.
Classify the document first to avoid unnecessary OCR and improve accuracy. This tool extracts text directly from PDFs that contain a text layer and runs OCR only for images or scanned PDF pages.
If you can select the words in a PDF with the mouse, it probably contains a text layer; direct extraction is usually faster and more accurate. A scanned page that behaves like one large image needs OCR.
Choose the recognition language that matches the main document language. High-quality mode helps with smaller text, while grayscale contrast enhancement can help faint monochrome scans. Enhancement may be less useful for normal color photos.
OCR renders pages as images and runs a language model, so it can use substantial memory. File-count, size, and PDF-page limits adjust to the current device; splitting a large PDF into smaller batches is often more reliable.
No. The original images and PDFs you select are not uploaded to the server; they are processed in your current browser's memory.
OCR calculations run on your PC or mobile device using PaddleOCR.js and WebAssembly. The MISCLAND server only delivers static OCR engine and model files; it does not receive or store document contents or OCR results.
If a PDF page already contains selectable text, PDF.js extracts that text directly. This is faster than OCR and preserves the original text more accurately.
Only scanned pages or image-based PDFs without a text layer are rendered as images in your browser and then processed with OCR. If a PDF contains both text pages and scanned pages, each page is handled automatically using the appropriate method.
You can select Korean, English, Spanish, Portuguese, French, German, Italian, Indonesian, Vietnamese, Arabic, Hindi, Simplified or Traditional Chinese, and Japanese.
The tool uses a PP-OCRv5 mobile recognition model suited to the selected language. The model is loaded only after you choose a language and actually start OCR, and browser caching is used for subsequent runs.
OCR cannot guarantee 100% accuracy because recognition can vary with source quality, text size, skew, background, and font.
Clear printed documents and scans produce the best results. Use High Quality for very small text, and try Document Enhancement for faint black-and-white scans. Always compare important results with the original document.
Yes. It works in modern mobile browsers that support WebAssembly and Web Workers.
OCR uses your device's memory and processing power, so low-spec devices or very large documents may take longer. On mobile devices, lower safety limits are automatically applied to file count, total size, PDF pages, and image resolution.