Bubble Blaster is a python script that remove text from speech bubbles in mangas/manhwas with OCR.
Install python (Make sure your python version is NOT more than 3.11):
https://www.python.org/downloads/
Download the latest release and unzip it.
https://github.com/Aeonss/BubbleBlaster/releases/latest/
If you have a compatible GPU, install CUDA for faster OCRs.
https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html
Run the "BubbleBlaster.exe" file included in the release.
To install manually, git clone the repository, and run bubbleblaster.py
git clone https://github.com/Aeonss/BubbleBlaster
- Click on the "Import Image(s)" button to select the images you want to OCR (remove the text of)
- You can select 1 image, or multiple by ctrl + clicking the images or drag selecting them
- Select the language that the text is in
- By default, the language is Korean, but Chinese, Japanese, and English is also supported.
- Please request for more languages here
- Select the confidence level by dragging the slider
- Confidence level is how "sure" the AI is that the text is the language you selected
- By default, the confidence level is 0.4.
- The higher the number, the less likely text is recognized and OCR'd, but the text is more likely to be the selected language
- The lower the number, the more likely that more text is recognized and OCR'd, but the text may include symbols or parts of the art
- Select any additional options
- "Export raw text" option will give a text file (.txt) of the text that the program OCR'd
- "Export translated text" option will give a text file (.txt) of the translated text that the program OCR'd
- "Preview image before exporting" option will give a preview window of all found text in the image along with the confidence level. This is so that you can play with the confidence level to get all the text you want to be removed.
- BubbleBlaster was created with EasyOCR and Deep-Translator.
- Please request any features or report any bugs in issues.
- Update icon
BubbleBlaster is released under the MIT license.