Project address: https://github.com/dmMaze/BallonsTranslator
One-click machine translation
The translation backfill refers to the estimation of the original text layout, including color, outline, angle, orientation, alignment, etc. The final effect depends on the overall performance of the four modules: text detection, recognition, erasure, and machine translation. Supports Japanese manga and American comics. The layout for English to Chinese and Japanese to English translations has been optimized, with text layout based on the extracted background bubbles. Chinese segmentation is based on pkuseg, and vertical layout for Japanese to Chinese translation needs improvement.
Image editing
Supports mask editing and repair brush.
Text editing
Supports WYSIWYG rich text editing and some basic formatting adjustments, font style presets. Supports find and replace for full text/source text/translated text. Supports import and export of Word documents. Suitable for comic strips.
Usage instructions
Windows
If you are using Windows and do not want to manually configure the environment, and can access the internet normally: Download BallonsTranslator_dev_src_with_gitpython.7z from MEGA or Google Drive, unzip it, and run launch_win.bat to start the program. If the libraries and models cannot be downloaded automatically, manually download data and ballontrans_pylibs_win.7z and unzip them into the program directory. Run scripts/local_gitpull.bat to get updates.
Running the source code
Install Python < 3.12 (do not use the Microsoft Store version) and Git.
# Clone the repository
$ git clone https://github.com/dmMaze/BallonsTranslator.git ; cd BallonsTranslator
# Start the program
$ python3 launch.py
The first run will automatically install dependencies like torch and download the required models and files. If the model download fails, you need to manually download the data folder from MEGA or Google Drive (or the missing files mentioned in the error) and save it to the corresponding location in the source code directory.