aurextranslator

AurexTranslator is a cross-platform, real-time translator for Windows and Linux (X11/XWayland) that translates text taken from the screen, the clipboard, or directly from a running game or engine. Its typical use case is overlaying translations on visual novels and other games, but the OCR and clipboard modes work with any on-screen text. Input sources - OCR: captures a screen region and recognizes text with Tesseract or Ollama Vision, in either automatic or manual capture mode. - Clipboard: translates copied text automatically. Depending on the desktop environment it can use a native Wayland backend, the XDG Portal, or the Qt clipboard backend. - Game text extraction: intercepts text output at runtime by injecting a plugin into the game's process and hooking its text output functions. This requires installing the injector (libat-injector plus at-injector) and a game- or engine-specific plugin; supported engines, games and install steps are listed in the separate AurexTranslator-plugins repository. Configuration and output - Named configuration profiles allow separate settings per game and let users switch between them without re-injecting. - Text replacement rules support regular expressions and can be scoped per source or per game. - Output can be revealed as a whole block or character by character with adjustable speed. Several text blocks can be selected with a configurable priority order, which helps with games or engines that expose multiple simultaneous text sources. - Multi-architecture plugin loading picks the correct x86 or x64 plugin build automatically. Platform support Windows and Linux (X11/XWayland). Native Wayland clipboard support is optional at build time; without it the application still builds and falls back to the XDG Portal and Qt clipboard backends. Roadmap (planned, not yet available) - Multi-zone OCR: capture and translate several screen regions at once instead of one zone. - Native Linux packages: .deb for Ubuntu/Debian, AUR for Arch, .rpm for Fedora/openSUSE, and Flatpak/Snap as universal formats. - In-game overlay: on-screen translation without a separate window, via graphics API hooking (DLL injection on Windows, LD_PRELOAD on Linux). - Text-to-speech voice output. - Additional translation engine providers. - More game and engine plugins. Building from source Common requirements are CMake 3.10+, Qt6 (Core, Network, Widgets, Svg, LinguistTools), OpenCV (core, imgproc, imgcodecs) and Tesseract OCR. Linux additionally needs Qt6 DBus, libpipewire-0.3-dev, libX11-dev, libXcomposite-dev, libXfixes-dev and libXrandr-dev; libwayland-dev is optional and enables the native Wayland clipboard backend. The documented build flow is: clone the repository, create a build directory, run cmake .. to configure, then cmake --build . to compile. Licensing The project's code is released under the GPL-3 license; the repository's LICENSE file holds the full terms. Feature requests and pull requests are directed to the project's issue tracker and pull request pages on GitHub.