Kimochi and friends
There are a few good tools for reading Japanese games. Here’s how they line up, so you can pick whatever suits you.
We tried to keep this fair. Everything below comes from each project’s own docs (see Sources), and anything we couldn’t confirm is tagged . Spot something wrong? Open an issue and we’ll fix it.
Kimochi owes a lot to owocr by Aurora Wright. Some of its OCR ideas came from there, it can take owocr’s output over WebSocket, and its Windows OneOCR support builds on oneocr. GSM runs a fork of owocr too.
At a glance
| Feature | Kimochi | Kamui | GSM | owocr | YomiNinja |
|---|---|---|---|---|---|
| Platforms | macOS 14+ (most mature), Windows, Linux. Steam Deck Gaming Mode is in early testing. | Web app in the browser, plus a Windows-only desktop app marked experimental. | Windows, macOS, Linux. Steam Deck in Desktop Mode (Game Mode not supported), or streamed from a PC. | Windows, macOS, Linux (Python 3.11 to 3.14). Linux works best on Wayland; X11 is partial. | Windows 10/11, macOS (Intel and Apple silicon), Linux x64. Built for X11; Wayland support is initial. |
| Capture method | Built-in screen or window capture with a crop region. No OBS needed. | Desktop app captures the game screen; the capture method is not documented. | OBS Studio (required, also used for screenshots and audio); OCR can also use native screen capture. | Screen area or window capture, OBS, or images from the clipboard, a folder or a socket. | The whole screen or one window, with OCR Templates for fixed text regions. |
| OCR engines | Google Lens (the default), Apple Live Text (macOS), MeikiOCR, Manga OCR, OneOCR (Windows), or a custom endpoint. | Google Cloud Vision (cloud). | A fork of owocr. Local: OneOCR, Screen AI, MeikiOCR, Apple Live Text, local LLM, MLKit. Cloud: Google Lens, Bing, Gemini, Google Vision, Azure, OCRSpace. | Local: Chrome Screen AI, OneOCR, Apple Live Text, Apple Vision, meikiocr, NDLOCR-Lite, Manga OCR, WinRT, EasyOCR, RapidOCR. Cloud: Google Lens, Bing, Google Vision, Azure, OCRSpace. | PaddleOCR, Manga OCR, Apple Vision (macOS), Google Lens, Google Cloud Vision. |
| Overlay on the game | Yes: recognised text drawn over the game, selectable for lookups (not with Manga OCR or owocr input, which return text only). | Yes, in the desktop app (experimental). | Yes: hover lookups with Yomitan, in-game card creation, optional furigana. | None documented; owocr outputs text for other tools. | Yes: extracted text over the screen for pop-up dictionary lookups; optional furigana filter; experimental browser overlay. |
| Texthooker, clipboard, WebSocket input | Clipboard and WebSocket text inputs alongside OCR; can also send OCR text to the clipboard or a WebSocket. | Not documented. | Yes: hooks from Agent, Textractor or LunaTranslator, plus clipboard; built-in texthooker page. | Takes images in (clipboard, WebSocket, folder); sends text out to the clipboard, a WebSocket or a file. | WebSocket for texthooker pages (sends OCR text out). Clipboard and text input not documented. |
| Dictionary lookup | Built-in Yomitan with dictionaries you import or download. | Built-in dictionary sidebar with context-aware lookups; AI translation. | Yomitan (in the overlay and texthooker); Jiten Reader in the overlay. | None built in; usually paired with a texthooker page and Yomitan. | Browser-extension dictionaries built in: Yomitan and 10ten pre-installed, other Chrome extensions partly supported. |
| Anki integration | Yes, through AnkiConnect. | Yes: one-click add from a dictionary entry, to your deck or a Kamui deck. | Its focus: enriches cards with the voiced line (trimmed by voice detection), screenshots or GIFs, and optional AI translation. | None built in. | Not built in; listed as a planned feature. |
| Automatic / continuous OCR | Yes: Auto OCR with change detection, optional local gating and adaptive polling. | Yes, in the desktop app (experimental). | Yes: two-pass, a fast local scan then a more accurate scan once text settles. | Yes: continuous scanning, repeated-text filtering, optional two-pass. | Yes: Auto OCR re-reads an OCR Template region when it changes (experimental, one region at a time). |
| Controller and handheld | Gamepad buttons drive the overlay menu and word lookups; Steam Deck Gaming Mode support is early. | Not documented. | Experimental gamepad navigation of the overlay; Deck in Desktop Mode. | Not documented. | Not documented. |
| Price and licence | Free. GPL-3.0. | Paid subscription ($5/month shown; yearly and lifetime options). Closed source. Users report a limited free allowance. | Free. GPL-3.0; accepts sponsorship. | Free. GPL-3.0. | Free. GPL-3.0. |
| Project status | Early: nightly pre-release builds, one developer. | Commercial product from Kamui, LLC; desktop app labelled experimental. | Active: release v2026.10.0 on 2026-10-04. | Active: release 1.26.8 on 2026-03-28, commits into June 2026. | Latest release v0.9.3 on 2025-08-04; no commits since. |
Checked . These projects move fast, so check their pages for the latest.
Which one?
Game Sentence Miner, if you live in Anki
GSM is all about great flashcards: the voiced line trimmed to fit, screenshots or GIFs, AI translations with your own key. It takes text hooks as well as OCR, has loads of engines, an overlay with Yomitan, and a big, active community. It needs OBS, which is more setup, but if you want a deck full of sentence cards with audio, it does more of that than Kimochi.
owocr, if you like building your own setup
owocr just does OCR, and does it well. It watches the screen, a window, OBS, the clipboard or a folder, runs one of many engines, and sends the text wherever you want. Great if you already have a texthooker page and Yomitan in the browser. Biggest engine list of the lot.
YomiNinja, if you want a free overlay with Yomitan or 10ten
YomiNinja puts the text it reads over the screen, so you can hover it with Yomitan or 10ten, which come pre-installed. It runs on Windows, macOS and Linux, with PaddleOCR, Manga OCR and Apple Vision on your machine or Google Lens and Cloud Vision online. Auto OCR can watch a dialogue box, and a WebSocket feeds texthooker pages. It reads Chinese and Korean too. Anki integration is still on its planned list, and its latest release is from August 2025.
Kamui, if you want it all handled for you
Kamui is a paid app, on the web and Windows. Quick to set up, no API keys, solid cloud OCR, a colour-coded sentence breakdown, AI translation and one-click Anki cards. You pay a subscription and your captures go to the cloud, but there’s a company behind it.
Kimochi, if you want it all in one app
Capture, OCR, the overlay, a log, Yomitan, Anki and a notebook, in a few small windows. No OBS, no browser tabs. Local engines or Google Lens, and a controller works too. It’s free and open source, but it’s young: no proper release yet, Steam Deck support is early, and the Anki side is simpler than GSM’s. More on the home page, or see what leaves your machine in the privacy policy.
Sources
- owocr: README (AuroraWright/owocr)
- owocr: releases
- Game Sentence Miner: README
- Game Sentence Miner: overview
- Game Sentence Miner: OCR
- Game Sentence Miner: overlay gamepad navigation
- Game Sentence Miner: Steam Deck
- Game Sentence Miner: releases
- Kamui: home page and FAQ
- Kamui: pricing
- Kamui: Nookgaming review
- Kamui: Bunpro community thread (free allowance report)
- YomiNinja: README (matt-m-o/YomiNinja)
- YomiNinja: releases (v0.9.3, v0.9.1 Wayland, v0.8.3 overlay, v0.7.2 Auto OCR)
- Kimochi: source
Kamui, Game Sentence Miner, owocr and YomiNinja belong to their own authors and aren’t affiliated with Kimochi.