Practical overview
When it fits
Use it when you need printed Japanese text extracted from manga or another image, especially where text direction, furigana, artwork overlays or varied fonts complicate recognition.
Who it is for
- Readers who want to look up Japanese manga text in a dictionary.
- Developers who need printed Japanese OCR for images in Python.
- People processing Japanese text in novels, games or other printed material.
Documented uses
- Extract text from manga speech bubbles without first separating each line.
- Recognize Japanese text in vertical or horizontal layouts, including furigana and text overlaid on artwork.
- Send captured image text to a clipboard-based dictionary workflow.
- Process an image directly through a Python interface.
Requirements & limits
The repository identifies the project as open source under the Apache-2.0 license. The README requires Python 3.9 or newer; GPU use is optional.
- A single recognition pass can handle multi-line text, but longer regions are more prone to errors.
- The model can be used for general printed Japanese OCR, though its main focus is manga.
- Clipboard scanning replaces copied images with recognized text; folder scanning is the documented alternative when preserving image copying matters.
- The model attempts recognition even on images without text, so it may return plausible but incorrect text.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Manga Ocr and recognize an image in Python
Set up the package and pass a local image to the Python interface for text recognition.
Before you start
- Python 3.9 or newer
- GPU and GPU-specific PyTorch setup are optional
- An image containing printed Japanese text
Steps
Install the package with pip install manga-ocr or uv add manga-ocr.
In Python, import MangaOcr from manga_ocr and create an instance with MangaOcr().
Pass an image path to the instance, or open the image with PIL.Image and pass the image object.
Expected outcome
The interface returns recognized text from the supplied image.
Guide limitations
- The newest Python release may not yet be supported because of PyTorch dependencies.
- Longer text regions are more likely to contain errors; try a smaller image region if recognition fails.
- The model may generate plausible-looking text when no text is present and is not intended for handwriting.
Set up clipboard-based manga OCR
Configure a documented workflow that sends captured manga text to a dictionary through the clipboard.
Before you start
- Python 3.9 or newer and Manga Ocr installed
- A screen-capture tool such as ShareX or Flameshot
- On Linux, wl-copy for Wayland or xclip for X11 clipboard mode
- A dictionary that can read clipboard text, such as Yomitan
Steps
Install Manga Ocr and start clipboard scanning by running manga_ocr.
Capture a manga region with the screen-capture tool and send the image to the clipboard.
Let Manga Ocr replace the clipboard image with recognized text.
Read the resulting clipboard text with the dictionary.
Expected outcome
Captured printed Japanese text is recognized and made available in the clipboard for dictionary lookup.
Guide limitations
- In clipboard scanning mode, any image copied to the clipboard is processed and replaced by recognized text.
- The first run downloads a model of about 400 MB.
- The project recommends folder scanning instead when preserving ordinary image copy-and-paste behavior is important.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- kha-white/manga-ocr · Official homepage title · retained in its source language
Related options
Poricom — A GUI reader that uses Manga Ocr.
mokuro — A tool that uses Manga Ocr to generate an HTML overlay for manga.