Free & Open Source · OCR Overlay · v0.9.3
YomiNinja captures on-screen Japanese text from games, visual novels, emulators, manga readers, and subtitled video, overlays it directly onto the window, and connects it to Yomitan's popup dictionary — so you can hover any kanji and get instant definitions without ever alt-tabbing.
The tool
YomiNinja reads on-screen Japanese text using optical character recognition, renders an interactive overlay directly on top of your game or application, and connects it to Yomitan's popup dictionary engine so you can hover any word, see the reading, and keep playing.
It works on any application where Japanese text is visible on screen — JRPGs, visual novels, emulators, manga readers, and video players with hardcoded subtitles. No game-engine hook list. No process injection. No need to move text into a browser first.
See how it works →The workflow
From pressing a hotkey to reading a definition in under a second.
Point YomiNinja at a game, visual novel, or any application window. Draw an OCR template over the dialogue box to target just the text region — or let Auto OCR scan the full window automatically.
Press the hotkey — or let Auto OCR fire when it detects new on-screen content. YomiNinja processes the capture through your chosen OCR engine and renders the extracted Japanese text as a transparent, interactive overlay on top of your game.
Move your cursor over any character in the overlay. Yomitan — pre-installed, no browser setup required — triggers instantly with readings, definitions, pitch accent, and frequency data. Your immersion stays unbroken.
Capabilities
The core app, dictionary lookups, OCR engine choice, and mining workflow all live in one stack.
No text hooking, no engine compatibility lists. If Japanese text is visible on screen, YomiNinja reads it — including games, emulators, visual novels with hardcoded or image-rendered text, manga readers, and video players.
Compatibility →Both dictionary extensions ship pre-installed. Import JMdict once, then start hovering kanji immediately. No browser, no Chrome Web Store, no separate extension install.
Dictionary setup →PaddleOCR for most games, MangaOCR for stylized fonts, Google Cloud Vision and Google Lens for hard cases, and Apple Vision Framework on macOS for strong native OCR including vertical text.
Compare engines →Continuous text detection that fires automatically when screen content changes. No hotkey required mid-battle. Define OCR templates over dialogue boxes to reduce OCR noise and improve speed.
All features →Hear extracted text read aloud using VOICEVOX, a high-quality Japanese speech engine. Built in as of v0.9.1 across all platforms. Listen while you read — reinforce pronunciation naturally.
Audio features →WebSocket output streams recognized text to any texthooker page in real time. Pipe sentences into your Anki sentence mining workflow without leaving the game. Direct Anki integration is still planned, so the current workflow uses WebSocket plus Yomitan.
Mining guide →Who it's for
Playing untranslated JRPGs, action games, and everything that never got a localization. YomiNinja turns Japanese games from inaccessible to learnable — one dialogue box at a time.
Working through visual novels and manga where text hooks don't reach. Hardcoded subtitles, image-rendered text, web-based readers — OCR captures what Textractor cannot.
Building Anki decks from native content, one sentence at a time. WebSocket output feeds directly into texthooker pages. Every sentence you read becomes a potential flashcard.
Availability
Windows 10 & 11. Full feature set including all OCR engines, VOICEVOX TTS, and GPU acceleration.
AppImage, .deb, .rpm, and .pacman packages. X11 required — Wayland experimental support added in v0.9.1.
Intel x64 and Apple Silicon (M1/M2/M3). Apple Vision Framework for native OCR including vertical Japanese text.
xattr -cr YomiNinja.appReality check
For games and emulators: OCR works where text hooks fail, especially on JRPGs, console ports, emulator windows, and titles with no usable hook.
For actual study: the main workflow is OCR overlay plus Yomitan popup, with optional WebSocket output for sentence mining and VOICEVOX for audio reinforcement.
Get started
Open source. No account. No subscription. No telemetry.
Windows, Linux, macOS — with clear setup paths for Yomitan, OCR engine choice, and first lookup.