How to use
Drop images or click “Choose files” (multiple files and Ctrl+V paste are supported). “Try a sample” shows how it works.
Under “Recognition settings” choose the language (Japanese + English, English only, or vertical Japanese) and preprocessing (grayscale, contrast boost). If an image is sideways, use “Rotate 90°” on that image.
Press “Read text”. Only the first time, the recognition engine and language data (about 9 MB for Japanese + English) are loaded from this site, with progress shown in MB.
The recognized text appears per image. Fix any mistakes right there. Images that were read with low confidence are flagged.
Use “Copy” or “Save as .txt”. For several images you can save everything in one .txt with a heading per image.
What it does
Removes extra spaces in Japanese
The recognition engine often puts spaces between Japanese characters. Blanks between kanji, kana and full-width symbols are removed, while blanks between Latin words and digits stay.
Images stay on your device
Recognition runs in a Web Worker inside your browser, so the page stays responsive. The engine and language data are served from this site, and no external service is contacted.
Loads only what it needs
Nothing is loaded just by opening the page. Data is fetched the first time you press “Read text”, and the vertical-text data only if you choose “Japanese, vertical text”.
Preprocessing to help reading
Choose grayscale and contrast boost; images whose longest side exceeds 4000 px are scaled down before reading. Rotate each image in 90° steps.
Fix and export the result
The result is an editable text box with copy and .txt save. Several images can be combined into one file with a heading per image name.
Use cases
- Turn text in a screenshot or photo into text without retyping it
- Photograph a paper document or note so it can be searched and edited
- Copy the text of an invoice or notice into another document
- Get English or foreign-language text from a sign or manual into a form you can paste into a translator
Supported formats
Input
Output
Capability matrix
| Feature | Support | Notes |
|---|---|---|
| Reading Japanese (horizontal) and English text | Supported | — |
| Reading vertical Japanese text | Partial | Complex column layouts can change the reading order |
| Several images at once, paste | Supported | Up to 50 MB per image |
| Editing, copying, .txt save, one file for several images | Supported | — |
| Handwriting | Not supported | Even when something is read, errors are frequent |
| Low-resolution images, decorative fonts, text close to the background colour | Partial | Larger, straighter images read better |
| Keeping the original layout (columns, positions) | Not supported | Only text is output |
| Reproducing table structure | Not supported | Tables come out as text with broken rows and columns |
| Creating a searchable PDF | Not supported | This tool outputs text only |
| Languages other than Japanese and English | Not supported | — |
Limits & notes
- Up to 50 MB per image. Images whose longest side exceeds 4000 px are scaled down to 4000 px before reading. Images over 200 megapixels cannot be opened.
- Handwriting, low-resolution images, strongly decorated fonts and text close to the background colour may not be read well. Always check the result by eye.
- The original layout is not kept. Columns can come out in a different reading order, and tables lose their rows and columns.
- The first run needs the recognition engine and language data (about 9 MB for Japanese + English). Later runs are faster thanks to the browser cache; clearing the cache loads them again.
- Reading time depends on image size and your device. On phones, large images or many images take longer and can run out of memory.
- When the average confidence is below 60%, “The reading is uncertain” is shown. This is a guide, not a guarantee of correctness.
Privacy & safety
Loading, preprocessing, recognizing and saving all happen inside your browser. Your images and the recognized text are never sent to or stored on a server, and are not part of usage statistics. The recognition engine and language data are served from this site, and no external service is contacted.
FAQ
Are my images uploaded?
Why does the first run take longer?
Why are there no spaces between Japanese characters in the result?
Can it read vertical text?
Can it read handwriting?
What should I do if it says the reading is uncertain?
Related tools
QR Code & Barcode Reader
Drop an image or point your camera to read QR codes and barcodes. URLs are never opened automatically — you see the real destination domain and phishing warnings first. Everything runs on your device.
Image Compressor
Drop images to shrink them in bulk while keeping them sharp. Convert to WebP or AVIF, resize, and compare before and after. Your images are never uploaded.
Images to PDF
Drop images, put them in order and get one PDF. Choose page size, orientation and margins; phone photos are rotated correctly. Images are converted on your device and never uploaded.
Phone Photo to Scan PDF
Take a photo of a page or pick one. The corners are found automatically, the page is flattened to a straight-on view, shadows are removed, and pages are combined into one PDF. Photos never leave your device.
Compress PDF
Drop a PDF to cut its size. Only photos and scans are recompressed, so text and graphics stay crisp. Your PDF is processed on your device and never uploaded.
PDF Editor
Lay text (Japanese included), shapes, arrows, a highlighter, freehand drawing and images on top of a PDF, and fill in its form fields. Everything is written as PDF vectors and real text, never as a picture. Rewriting the original text and true redaction are not possible.
Last updated: