Ever tried learning a language and gotten frustrated because pop-up dictionaries only work inside your browser? Worry no more!
Lexiglance is a system-wide pop-up dictionary for Linux (X11) and Windows. Hold a key combination and point at a word in any application to see its definitions. It supports Japanese, Russian, Ukrainian, Korean and Greek, and uses the same dictionary format as Yomitan.
Download for Windows · Download for Linux · Other packages
- Reads text through the accessibility interfaces (AT-SPI, UI Automation), or with OCR (PaddleOCR, Tesseract) in games, images and videos.
- Dictionaries are compiled into memory-mapped indexes.
- Deinflection: 食べさせられなかった finds 食べる, книгами finds книга.
- Whole sentences translated into English, on your computer: hold the sentence key with the trigger, or select the text. The translation models (OPUS-MT, about 115 MB a language, or 440 MB in full precision, which translates about twice as fast - each language uses whichever you pick) run offline, so nothing you read is sent anywhere.
- Recommended dictionaries for each language can be installed from the settings application.
- Pronunciation audio, and Anki cards through AnkiConnect.
- Configurable popup designs, colour schemes and highlight styles (docs/themes.md).
- A health check that diagnoses common setup problems, with a quick fix button.
- Statistics: what you have looked up, in which languages and on which days, kept on your computer only.
- No input injection and no access to other processes, it won't trigger any anti-cheats - feel free to use it in games! (details).
- Windows 10 (1809 or later) and 11: download lexiglance-windows-setup.exe and run it. Administrator permissions are not needed, it may also warn about an unknown publisher: choose More info -> Run anyway. Running a newer one updates in place.
- Linux: download
lexiglance-x86_64.AppImage,
make it executable (
chmod +x lexiglance-x86_64.AppImage) and run it.
The releases page also has a portable .zip for Windows, a .deb
for Debian and Ubuntu, a .rpm for Fedora, and a .tar.gz. From then on Lexiglance keeps itself up to date
(Overview -> Updates), in the background too while you are away from the computer.
- Open Dictionaries -> Get recommended dictionaries, pick a language and install.
- Hold Super + Left Alt (on Windows Win + Left Alt) and point at a word.
- Keep holding the keys and use the scroll wheel to select more characters (or fewer). Left-click an entry to copy it. Right-click and select text within a pop up box to copy it instead.
- Hold Left Shift as well to see the sentence in English, or only the characters you selected. The Translation page downloads the models (compact or full precision, per language) and sets the key.
For games, videos and other programs that expose no text, click Scanning -> Download PaddleOCR once. On Linux,
Chromium and Electron apps need --force-renderer-accessibility.
Each language is a small JSON file, and you can add your own without rebuilding. See docs/languages.md.
You need CMake 3.30+, GCC 14 or Clang 18+, zlib, Cairo, Pango, libcurl and Qt 6.4+. On Linux also Xlib (with XInput2, XFixes, XRandR, Xext) and AT-SPI 2.
cmake --preset release
cmake --build --preset release
ctest --preset releaseThen run build/release/apps/gui/lexiglance. For Windows builds see docs/porting.md.
See CONTRIBUTING.md. Please report security problems privately, see SECURITY.md.
Yes, this one was mostly written with the help of an LLM. I wanted to get the program up and running as quickly as possible so that my friends and I could start using it in our language-learning journeys. Writing everything by hand would have taken a very, very long time!
Of course, I still review the code and documentation myself. I don't want to get completely lost in the madness
I hope Lexiglance can be a useful tool to others and help everyone on their own language-learning journey!
MIT LICENSE.
By MattFor
Dictionaries, OCR and translation models and recordings are downloaded separately and keep their own licenses. Lexiglance includes the ONNX Runtime headers (MIT), a kanji table from OpenCC (Apache 2.0) and Breeze icons (LGPL 3.0).


