Introducing ViewTranslate

North Wolf Labs is introducing ViewTranslate, a new project focused on making more software usable across languages.

The idea is straightforward: select an app, window, or display, and ViewTranslate translates visible text directly over the original content. Instead of switching between screenshots, copy-paste, browser tabs, or separate translation tools, the translated text stays where the user needs it: on top of the software they are trying to use.

This is not meant to replace proper localization. When an app is fully localized, that will usually be the best experience. It is also not meant to compete with browser translation, since the translation tools built into modern browsers are already quite good. ViewTranslate is for everything else: software that was never translated, tools with incomplete language support, games that are only available in certain regions, or workflows where language is the one thing getting in the way.

The project is still in its early stages, but the first Windows proof of concept is complete. At the moment, ViewTranslate exists as a developer-facing C++ command line tool that connects the core pieces together: frame capture, OCR, translation, and the early translation-overlay workflow.

Some of the foundation has already been chosen. On Windows, frame capture is built around Windows.Graphics.Capture. Local OCR starts with Windows OCR and PaddleOCR, with PaddleOCR running through ONNX Runtime for better CUDA support on Windows. Local translation is built around CTranslate2.

Early testing found that Windows OCR falls short on quality, while PaddleOCR can provide results comparable to Google Cloud Vision locally and in less time. CTranslate2 has provided decent translation quality at excellent speeds, but so far, local OCR paired with cloud translation has produced the best overall experience.

Cloud providers will be part of the first version too. Google Cloud Vision is the initial cloud OCR provider, while Google Translate and DeepL are the first planned cloud translation providers. More providers may be added later as the project develops.

The end goal is a native desktop app for Windows and macOS. Windows will use WinUI 3, and macOS will use SwiftUI. The product should feel like a practical desktop tool: choose what to translate, pick how it should be translated, and keep working.

ViewTranslate is in active development, with an expected release target of Q4 2026.