Manga Scan Translator (Chrome Extension Edition)

Translate manga or comic speech bubbles directly in your web browser. This project uses a Local Python Server (FastAPI) to process the images, ensuring your translations are unlimited, private, and free!

1. Setup the Python Backend

1. Open a terminal in this directory.
2. Install PyTorch with CUDA support (since you have an NVIDIA GPU, this will make OCR much faster): Visit PyTorch's official Get Started page to find the exact command for your CUDA version. For example, if you have CUDA 11.8:

   pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

3. Install the rest of the dependencies:

   pip install -r requirements.txt

4. Setup the Font:
   • Download a Thai .ttf font (e.g., 'Mali' or 'Itim' from Google Fonts).
   • Create a folder named font in this directory if it does not exist.
   • Place the downloaded .ttf file inside the font folder and name it thai_comic.ttf.
5. Start the Server:

   uvicorn server:app --port 8000

   Note: Leave this terminal window running in the background. The first time you start it, it may take a minute to load the OCR models into your GPU.

2. Install the Chrome Extension

1. Open Google Chrome and go to chrome://extensions/.
2. Toggle on "Developer mode" in the top right corner.
3. Click the "Load unpacked" button in the top left.
4. Select the extension folder located inside this project directory (d:\Work\Code\Manga Translator\extension).
5. Pin the "Local Scan Translator" extension to your Chrome toolbar for easy access!

3. How to Use

1. Make sure your Python uvicorn server is running.
2. Visit any website where you are reading manga or comics.
3. Click the Local Scan Translator icon in your Chrome toolbar.
4. Select your language (Japanese -> Thai OR English -> Thai).
5. Click "Start Snipping".
6. Your cursor will turn into a crosshair. Click and drag a red box tightly around a speech bubble on the page.
7. Wait a few seconds. The text will be extracted, translated, typeset, and injected back into the web page, perfectly overlaying the original bubble!