Translumo

Why we included this project

Translumo handles a genuinely annoying problem: text you cannot select or copy because it is burned into the image, like subtitles baked into a video or the dialogue in a game you play in a language you only half read. You mark a region of the screen, and the tool runs OCR on it, sends the result to a translation service, and shows the translation as an overlay within a short delay. The interesting part is how it gets accurate readings: it runs several OCR engines at once and uses a lightweight machine-learning model to pick the best result, which matters for stylized game fonts and text that flickers past. Because it works anywhere on the display, the same setup translates PDF readers, video players, and desktop apps without touching the original files. It runs natively on Windows, and you can pick among the built-in engines and services, from Windows OCR or Tesseract to DeepL, Google, or Yandex.