Screen Translate

A Python tool to capture a selected screen region on GNOME, extract text using Tesseract OCR (French by default), and translate it using either OpenAI GPT or Google Translate API.
Includes a Tkinter GUI for easy use and editable translation results.

---

Features

• Region capture using gnome-screenshot
• Image preprocessing for improved OCR accuracy
• OCR with Tesseract (default language French)
• Translation via OpenAI GPT or Google Translate API
• Editable original text with re-translation option
• DPI scaling support based on monitor config

---

Installation

Make sure you have the required system dependencies and Python packages installed.

System dependencies (Debian/Ubuntu)

sudo apt update
sudo apt install python3-pip python3.13-venv python3-tk gnome-screenshot tesseract-ocr tesseract-ocr-fra

• Replace tesseract-ocr-fra with your source language pack if needed (see notes below).

Python environment setup

python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install \
    pillow \
    pytesseract \
    opencv-python-headless \
    numpy \
    requests \
    python-dotenv \
    openai \
    tk

---

Configuration

.env file

Create a .env file in the project root with your API keys:

OPENAI_API_KEY=xxx-your-openai-key-xxx
GOOGLE_API_KEY=xxx-your-google-translate-key-xxx

Make sure to replace the placeholders with your actual API keys.

---

Language settings

By default, the script is set to:

SOURCE_LANG = 'fra'  # Tesseract language code (e.g. 'fra' for French, 'eng' for English)
TARGET_LANG_OPEN_AI = "English"  # Target language for OpenAI GPT translation
TARGET_LANG_GOOGLE_TRANSLATE = "en"  # Target language for Google Translate API

You can change these variables in the script to your preferred source and target languages. Examples:

SOURCE_LANG = 'eng'  # English OCR
TARGET_LANG_OPEN_AI = "Spanish"
TARGET_LANG_GOOGLE_TRANSLATE = "es"

Make sure you have installed the corresponding Tesseract language pack, e.g.:

sudo apt install tesseract-ocr-eng

---

OpenAI Model & Translation Prompt

The current OpenAI translation uses the gpt-4o model with a prompt designed for comic-style text preserving tone and humor:

model="gpt-4o"

prompt = (
    f"Translate the following comic-style text to {target_lang}. Preserve any tone, humor, or emotion.\n\n"
    f"Text:\n{text}\n\n"
    f"Translation:"
)

Feel free to adjust the model or prompt in the translate_with_chatgpt() function to suit your use case.

---

Usage

Run the script:

python3 screen_translate.py

• Select your preferred translator in the GUI dropdown.
• Click Capture & Translate and select the screen region.
• The extracted text and translation will appear in a popup window.
  • You can edit the OCR-extracted text in the top field if corrections are needed.
  • Then click 🔄 Retranslate to get an updated translation based on your edits.

---

Example

Below is an example of the GUI in action:

---

Notes

• This tool currently works on GNOME/Linux environments due to reliance on gnome-screenshot and .config/monitors.xml for DPI scaling.
• If you want to add support for other desktop environments or OS, adapt the screenshot method and scaling retrieval accordingly.
• Error handling is basic — check console output for issues.
• The code was generated and refined with the help of ChatGPT.

---

License

MIT License

---

Acknowledgements

• Tesseract OCR
• OpenAI API
• Google Translate API
• ChatGPT for code generation assistance

🤖 Generated with ChatGPT This codebase was generated and refined using ChatGPT-4 to speed up prototyping and development.