Kanji Reader uses the latest in convolutional neural networks technology to create an image processing application which converts photos of individual Japanese characters, providing their readings, translations, and other information.

Simply aim the camera at a character, and watch the results display live. No internet connection is required. Recognizes with 90+% accuracy over 2000 Kanji.

Toggle between white and black text detection modes. Adjust camera sensitivity to pick up fainter text, or reduce sensitivity in high contrast environments.

View more results returned by the model, up to 10, ordered by likelihood.

View detailed information about each Kanji, such as the "On" and "Kun" readings, radical, and stroke order.