JitenOCR BROWSER LAB
v8 · local inference

JAPANESE OCR, ON YOUR DEVICE

A page in.
Japanese text out.

Try v8 on a Kingdom page or your own image.
Images stay in your browser. No OCR server.

Checking WebGPU…

Desktop and mobile browser prototype

Loading sample pages…

First run downloads about 80 MB of model weights plus browser runtimes. Models are cached when browser storage is available. Uses greedy decoding.

PAGE TIME—Run a page to measure
DETECTION—Resize + model + boxes
RECOGNITION—Recognizer + text decoding
TEXT LINES—Both text orientations

Source page

Drop an image here

Recognized text 認識結果

Ready when you are.

Read a page to see its text here.
Tap a line to highlight it in the image.

Run details
Timings and device information appear after a run.