Image Text Recognition (OCR)
Processing entirely on your device — large files may take a moment.
Printed a form, photographed a page, or received a scan that you need as editable text? OCR (optical character recognition) reads the characters out of an image and hands them back as plain text. This page runs Tesseract — the open-source OCR engine with thirty years of history — directly in your browser: pick English, Simplified Chinese, or both, drop in JPG/PNG/WebP/BMP/GIF/TIFF images, and the recognized text appears with a copyable result.
Nothing is uploaded. The recognition engine and its language data download once into your browser and stay cached, so after the first run the tool even works offline. For the searchable-PDF output, the recognized text is layered invisibly over your original image — the result looks like the scan but selects, searches and copies like a text document.
How to extract text from images (OCR)
- 1
Select your files
Drag & drop your image files onto the area above, or click to browse. You can add multiple files at once for batch conversion.
- 2
Convert instantly
Click "Recognize text". The OCR engine runs locally on your device — the language data downloads once and stays cached in your browser.
- 3
Download results
Download each converted TXT file individually, or grab everything at once as a ZIP archive.
About the JPG format
JPG (short for JPEG, the Joint Photographic Experts Group committee that standardized it in 1992) has been the default photo format for over three decades. It compresses photographs using the discrete cosine transform: the image is split into 8×8 blocks, and details the human eye is bad at noticing are discarded. The result is files many times smaller than uncompressed data, at a quality level you control when saving.
In practice JPG is unbeatable for photographs and smooth gradients — camera memory cards, messaging apps and the web all assume it. It is a poor fit for screenshots, logos or anything with sharp edges and flat color, where compression leaves visible ringing around text. There is no transparency support, color depth is capped at 8 bits per channel, and every re-save with compression compounds tiny generational losses.
+ Strengths
- ✓Opens on literally every device, app and browser made in the last 30 years
- ✓Compresses photos dramatically — typically 10:1 or better with little visible loss
- ✓Quality level is adjustable, so you can trade fidelity for file size
- ✓Mature ecosystem: EXIF metadata, ICC profiles and progressive loading all supported
! Limitations
- −Lossy — artifacts appear on text, sharp edges and flat areas
- −No transparency, so overlays and logos need PNG or WebP instead
- −Repeated re-saving at low quality permanently degrades the image
About the TXT format
TXT is the floor of computing: a sequence of characters and nothing else. No headings, no bold, no embedded images — and because there is nothing to interpret, every editor, terminal, programming language and operating system ever built opens one. Modern files are UTF-8; older ones may be GBK, Latin-1 or Shift-JIS, which is the only surprise plain text can still deliver.
Its limits are obvious — no formatting means no structure — yet that is exactly why TXT refuses to die: logs, notes, code, READMEs, config fragments and data dumps all live happily as plain text. When you need the content and not the styling, TXT is the lowest-friction answer there is.
+ Strengths
- ✓Opens absolutely everywhere, including terminals
- ✓Tiny files, no parser, no corruption vectors
- ✓Perfect for logs, notes and machine-to-machine text
! Limitations
- −No formatting, structure or media whatsoever
- −Legacy encodings can garble non-English text
Frequently asked questions
Which languages can it recognize?
English and Simplified Chinese are bundled, and you can run both together for mixed documents. The language data downloads once and is cached by your browser.
Are my images uploaded to a server?
No. The OCR engine runs entirely inside your browser via WebAssembly — your images never leave the device.
How do I get the best accuracy?
Use the sharpest source you have: 300 DPI scans, straight crops, even lighting and clean backgrounds dramatically improve results. For screenshots, avoid scaling artifacts.
What is the searchable PDF output?
It places the recognized text as an invisible layer over your original image: the page looks identical to the scan, but the text is selectable, searchable and copyable in any PDF reader.
Does it work on handwritten notes?
Tesseract is tuned for printed text. Neat print handwriting works sometimes; cursive handwriting does not — for that, dedicated handwriting recognition is the right tool.
Is there a size or page limit?
The practical limit is your device memory. Process images one at a time or as a batch — multi-page TIFFs and long screenshots both work.