ramjke/Translumo

Advanced real-time screen translator for games, hardcoded subtitles in videos, static text and etc.

Translumo – Real‑time screen translator

Translumo is a Windows desktop application that captures a user‑defined region of the screen, runs OCR on the captured image, translates the recognised text and overlays the result back onto the screen. It is aimed at gamers and anyone who needs fast, on‑the‑fly translation of on‑screen text.

What it does

  • Capture a rectangular area (hotkey Alt+Q) from any window that is currently active.
  • Run OCR using one or more engines (Windows OCR, Tesseract 5.2, EasyOCR). The app scores the outputs with a small ML model and picks the best result.
  • Translate the recognised text via an online service (DeepL, Google Translate, Yandex, Naver Papago).
  • Display the translated text as an overlay; hotkey ~ toggles the translation view, Alt+T hides/shows the overlay.

Key features

Feature Details
High OCR precision Combines multiple OCR engines and selects the best output using a scoring model.
Low latency Optimised pipeline and optional GPU‑accelerated EasyOCR for faster processing.
Game‑friendly Works in borderless or windowed mode; overlay stays on top of the game.
Multiple translators DeepL (default), Google, Yandex, Papago – selectable per session.
Broad language support OCR: EN, RU, JA, ZH‑CN, KO. Translation: 30+ languages including French, Spanish, Arabic, etc.
Proxy support Allows rotating 1‑2 IPv4 proxies to avoid rate‑limits from translation APIs.
Customisable hotkeys All main actions are bound to keyboard shortcuts.

System requirements

  • Windows 10 (2004+) or Windows 11
  • DirectX 11‑compatible GPU (required for the Windows OCR path and optional EasyOCR)
  • 2 GB RAM for the basic Windows‑OCR/Tesseract mode; 8 GB RAM + NVIDIA GPU with CUDA 11.8 for EasyOCR
  • ~5 GB free disk space (the installer downloads OCR/translation models ≈ 400 MB)

Getting started (user side)

  1. Download the latest zip from the releases page and run Translumo.exe.
  2. Press Alt+G → Settings → choose source OCR language and target translation language.
  3. Pick the OCR engine(s) you want (Windows OCR is recommended).
  4. Press Alt+Q, drag a rectangle over the text you want to translate.
  5. Press ~ to start translating; the overlay appears on the screen.
  6. Use the Proxy tab if you hit API rate limits.

Building from source

  • Prerequisites: Visual Studio 2022, .NET 8 SDK.
  • Clone the repo, open the solution, and build.
  • The binaries_extract.bat script runs automatically during the build to download the required Python binaries and model files (~400 MB).

Credits & dependencies

  • UI: Material Design In XAML Toolkit
  • OCR wrappers: Tesseract .NET, EasyOCR (via Python.NET)
  • Image handling: OpenCvSharp
  • TTS (optional): Silero models

License

Apache 2.0 – free to use, modify, and distribute.


In a nutshell: Translumo bundles OCR, a lightweight ML scorer, and online translation APIs into a single Windows tool that lets you translate any on‑screen text in near‑real time, with a focus on low latency for gaming scenarios.

Related

  • Project
  • Project
  • Project
  • Project