Kotoba Snap

Snap a photo. Turn the Japanese words in it into flashcards.

Menus, street signs, packaging, a page of a novel - whenever you run into Japanese you don't know, just photograph it. Kotoba Snap recognizes the text, splits it into words, and lets you pick the ones worth learning.

WHAT IT DOES

- Photo to text. Handles vertical Japanese, and recognition runs on your device.
- Instant lookup. An offline dictionary gives readings, part of speech and JLPT level.
- Verb conjugation. Expand any verb into its full set of forms and derived usages.
- Spaced review. The schedule is handled for you - no deciding when to revisit a word.
- Source photos. Browse your cards grouped by the picture you found them in.
- Reference translation. Recognized sentences can be translated into Simplified Chinese on your device. On Android it routes through English, so the meaning is often off - treat it as a hint, not an answer.
- AI translation, optional and off by default. If you want a translation you can rely on, bring your own API key from Claude, ChatGPT or DeepSeek. See PRIVACY below.

INTERFACE LANGUAGES

English and Simplified Chinese. The app opens in Simplified Chinese - switch to English under Settings > Language. A Japanese interface is not available yet.

Definitions are provided in Simplified Chinese and English.

NO ACCOUNT, NO SIGN-UP

There is nothing to register for. The dictionary ships inside the app, so lookup and review work with no network at all.

PRIVACY

Your photos never leave your device, in any mode. The text recognized from them is not uploaded to us either, and Google's ML Kit terms state it is not sent to Google.

The one exception is AI translation, and it is off until you set it up yourself. With it on, translating sends the recognized text - only that text, never the photo - to the provider you picked, using an API key you obtained yourself. The provider bills you directly; we take no part in it and unlock nothing for payment. Your key is stored only on this device. You are asked for consent before the first request to each provider, and again if you switch.

The app is not entirely offline, and the privacy policy says so plainly: the bundled Google ML Kit SDK sends limited diagnostic and usage data. Full details, in plain language:
https://kotobasnap.github.io/privacy/en/

DICTIONARY DATA

Japanese dictionary data comes from the JMdict/EDICT project (Electronic Dictionary Research and Development Group), used under CC BY-SA 4.0. JLPT level tags are community estimates, not official. Simplified Chinese definitions are machine-generated and have not been reviewed by a human - full attribution is in the app under Settings > Third-Party Licenses.