owokuro: Generate Mokuro Files with OCR

owokuro

https://github.com/kamperemu/owokuro

The script allows you to generate mokuro files using owocr which allows you to use a variety of ocr engines that maybe faster and/or more accurate than manga-ocr + comic_text_detector.

Install and usage instructions in README.md.

A new version of owocr (1.26.3) was also released, which improves text separation/reordering and the furigana filter for vertical text. Older versions incorrectly merge bubbles much more often, so it's recommended to update.

The project does not have feature parity with mokuro and its a bit of a hacky proof of concept. It should work fine for most use cases but my plan is to properly integrate this directly into mokuro reader so I'm not going to further polish this. This interface is mostly to gauge user interest. Basically don't expect new features to be added by me (feel free to make pull requests though).

Regardless bug reports are appreciated.

Special thanks to Aurora Wright and Gnathonic for helping with some developer stuff

Install Instructions

Install uv (should theoretically work with python3-pip as well)

`

uv tool install git+https://github.com/kamperemu/owokuro

`

Usage Instructions

Install and open owocr with read_from = websocket, write_to = websocket, output_format = json (cli argument or config file/editor) and keep it open in the background.

If you're running owocr through cli you can run:

`

owocr -r websocket -w websocket -of json

`

If your directory structure looks somewhat like this:

manga/

├─vol1/

├─vol2/

├─vol3/

└─vol4/

You can process one volume by running:

`

owokuro /path/to/manga/vol1

`

You can process all volumes by running:

`

owokuro --parent_dir /path/to/manga

`

This will generate mokuro files for each volume in /path/to/manga folder which can be used with mokuro reader.