Comic Scans

A project to index a manga and make the entirety of its text searchable.

This project is powered by cubari.moe for the images. The OCR model used is GLM OCR, which is the most effective from our tests.

Setup

Ideally you should use a package manager to set this thing up like uv

uv venv --python 3.12
source .venv/bin/activate
uv pip install -qr requirements.txt

Usage

Run

python comic_ocr.py -i <manga-link>

Or you can skip the parameter and it'll ask you automatically.

Additional flags:

• --chapters 1 2 5 — only OCR the specified chapter numbers
• -f / --force — re-OCR chapters that are already present in the output JSON

Running on local images

If your image URLs inside the cubari JSON point to locally stored files, they must use the file:/// scheme. For example:

{
  "title": "My Manga",
  "chapters": {
    "1": {
      "groups": {
        "Group": [
          "file:///home/user/manga/ch1/page01.jpg",
          "file:///home/user/manga/ch1/page02.jpg"
        ]
      }
    }
  }
}

To use a local cubari-compatible JSON instead of a cubari link, pass --file:

python comic_ocr.py --file ./my_series.json
python comic_ocr.py --file ./my_series.json --chapters 1 5 10
python comic_ocr.py --file ./my_series.json -f

The JSON must have at minimum a title field and a chapters field following the cubari series format. --file and -i are mutually exclusive.

License

This project is licensed under MIT