kha-white/manga-ocr

Health v2 · maintenance only

Adoption not part of the score

2758 stars · 140 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Manga OCR is a Python library providing optical character recognition for Japanese text, focused on Japanese manga. It uses a custom end-to-end Vision Encoder Decoder model built with Hugging Face Transformers, robust to vertical/horizontal text, furigana, overlaid text, varied fonts, and low-quality images.

Use cases

• extract japanese text from manga panels
• ocr japanese text from images
• recognize text in manga speech bubbles
• read vertical japanese text from scans
• build a japanese reading and mining workflow
• process furigana-laden text in comics
• use as a general purpose printed japanese ocr

When to choose

• you need high-quality OCR for Japanese manga or printed Japanese text
• you need multi-line text recognition in a single pass
• your images are low quality, have furigana, or text overlaid on art
• you want a simple Python API or CLI to integrate into a reading workflow

When to avoid

• you need handwriting recognition rather than printed text
• you need OCR for languages other than Japanese
• you need scene text detection/localization rather than recognition of a given region
• you cannot run a PyTorch/Transformers model locally

Facets

library · maturity stable

ocr image-processing machine-learning deep-learning nlp computer-vision image-processing machine-learning python cross-platform japanese manga transformers vision-encoder-decoder furigana pytorch natural-language-processing

Member repositories

For agents

markdown · JSON · MCP: product_card(name="kha-white/manga-ocr")