Screen Translator

Press Ctrl+Alt+T to translate anything on screen

An async, GPU-accelerated Windows app that captures any on-screen text with OCR, translates it via Google Translate, and overlays the result at the original coordinates. Includes a full PyQt6 dashboard with GPU status indicator, Start/Stop controls, and live settings. No disk writes. No API key needed. Runs in your system tray.

Download Source Code~45 KB ZIP

Features

Dashboard GUI

Full PyQt6 dashboard with GPU status indicator (green for GPU, red for CPU), Start/Stop translation toggle, and live settings.

Instant OCR

EasyOCR with CUDA acceleration extracts text from the screen in under a second on supported GPUs.

Bugfix: Language Crash Fixed

v1.1.0 fixes the ValueError crash caused by mixing incompatible language groups (CJK + Latin) in the EasyOCR Reader. Now uses a safe ['en', 'tr'] default with automatic validation.

Auto-Translate Mode

Toggle from the dashboard or tray menu. Scans the screen every ~1.5 seconds and only re-translates when content changes.

GPU Status Indicator

Detects your NVIDIA GPU on startup (e.g. RTX 5070 Ti). Shows a green indicator when CUDA is active, or a clear red warning to run the CUDA fix command.

No Disk Writes

All image processing happens in memory via io.BytesIO. Nothing written to disk during operation.

How it works

  1. 1
    Capture

    Press Ctrl+Alt+T or click from the tray. A full-screen snapshot is taken via mss.

  2. 2
    Extract

    EasyOCR scans the image for text using GPU-accelerated deep learning models.

  3. 3
    Translate

    Extracted text is sent to Google Translate's free API (no key needed).

  4. 4
    Overlay

    Translated text is rendered at the original screen coordinates in a sleek overlay.

Requirements

OSWindows 10 / 11
Python3.10+
GPUNVIDIA RTX 20-series+ (optional - CUDA 12.x)
RAM4 GB (8 GB recommended)

Windows note: The global hotkey (Ctrl+Alt+T) uses the keyboard library. Running as Administrator is required for the global hotkey on most Windows configurations. If the hook fails, the app automatically falls back to polling mode (no admin needed). For the full hotkey experience, right-click your terminal and select "Run as administrator". To enable CUDA, install PyTorch with:pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

Ready to try it?

Download the source code, install the dependencies, and runpython main.py

Download Source Code