jedzqer/manga-translator-android

What it solves

This project provides an Android application that automates the translation of manga, allowing users to read translated versions of comics in their preferred language. It solves the problem of manually translating text within images, by automating the bubble detection, text recognition, and translation process.

How it works

The app uses a pipeline of local and API-based tools to process manga pages:

1. Bubble Detection: It uses a YOLO26n-seg model to detect and segment manga speech bubbles.
2. OCR (Optical Character Recognition): It recognizes text within those bubbles using either local models (like PaddleOCR) or OpenAI-compatible OCR APIs.
3. Translation: The recognized text is passed to a Large Language Model (LLM) via an OpenAI-compatible API for translation into the target language.
4. Overlay: The translated text is displayed as draggable bubbles that overlay the original image, preserving the original layout.

Who it’s for

Readers of manga and comics who want to read content in languages they are not fluent in, as well as users who want to translate screen content in real-time via a floating window translation feature.

Highlights

• Screen Translation: A floating window allows users to translate text from any app or screen on their Android device.
• Multi-language Support: Supports a wide range of source languages (Japanese, English, Korean, French, Spanish, Portuguese, German, Italian, Russian) and multiple target languages based on the app's interface language.
• Manga Library Management: Supports importing images, CBZ, ZIP, and PDF files, with automatic progress saving.
• Glossary Support: Maintains a glossary.json file per folder to ensure consistent translation of names and terms.
• Flexible OCR: Users can choose between local OCR models for privacy and speed, or API-based OCR for potentially higher accuracy.