Image to Text (OCR for English & Japanese)
Pick an image and get the text in it. Works on screenshots, photos of documents and scans. The OCR engine (Tesseract) runs inside your browser, so your image is never uploaded to a server. To make text in an image harder for OCR to read, try the OCR-Resistant Image Obfuscator.
🔒 Runs in your browser · Your data is never sent to a server
Accuracy is not guaranteed. Image quality, fonts and layout can cause mistakes or missing text, so always check important content against the original.
Click to choose, or drag and drop here (PNG, JPEG, WebP and more, up to 20 MB)
The first run downloads the OCR engine and language data (a few MB), so it takes a little longer. Later runs use the copy saved in your browser.
Confidence is low. Enlarging the image, raising its contrast or changing the language may help.
How to use
- Choose an image file (or drag and drop it)
- Pick the language to recognize (use "Japanese + English" if unsure)
- Press "Extract text"
- Check and fix the result, then copy it or save it as a text file
Notes
- Accuracy is not guaranteed. Small text, low resolution, tilted shots, handwriting, decorative fonts and vertical text produce more errors.
- Only on the first run, the OCR engine and language data (a few MB for Japanese and English) are loaded from this site. No external server is contacted.
- Images are processed only inside your browser. If the recognized text contains personal information, handle it with care.
- Table borders and layout are not reproduced; the recognized text is simply listed in reading order.
- Supported formats are PNG, JPEG, WebP, BMP and GIF (not HEIC or SVG). A recent browser with WebAssembly SIMD support is required.
Glossary
OCR
Short for Optical Character Recognition. Technology that turns text in an image into text data you can copy and search.
Tesseract
An open-source OCR engine. This tool runs Tesseract as WebAssembly so recognition happens inside your browser.
Confidence
A 0–100% guide to how certain the engine is about its result. A high value does not guarantee there are no mistakes.
FAQ
Is my image uploaded to a server?
No. Recognition runs inside your browser (WebAssembly), and neither the image nor the result is sent anywhere. The OCR engine and language data are also loaded from this site itself, not from an external CDN.
Why is the first run slower to start?
The OCR engine and language data (a few MB for English and Japanese) have to be loaded first. They are then saved in your browser, so later runs start much faster.
Can it read handwriting or vertical text?
Accuracy drops sharply. The tool is meant for printed, horizontal text; handwriting, vertical text, decorative fonts and small text produce more mistakes and gaps. Use an image with high resolution and clear contrast whenever you can.
What can I do if the result has many errors?
Enlarging the image, cropping to the text, straightening a tilted shot, or limiting the language to Japanese only or English only can help. Accuracy is never guaranteed, so check the result against the original.
Are tables and layout preserved?
No. Borders and columns are not reproduced; the recognized text is just listed in reading order. If you need a table, copy the result and format it separately.
Share this page