MangaTranslator is an AI-powered Gradio web application for automated manga and comic page translation. It targets speech bubbles and external text, offering a one-click solution across 59 languages to streamline visual content localization.

How It Works

The pipeline uses YOLO and SAM for text detection/segmentation, FLUX models or OpenCV for inpainting cleaned areas, and LLMs/VLMs for OCR and translation across 59 languages. Translated text is rendered with custom fonts, and output images can be upscaled via 2x-AnimeSharpV4. It supports single/batch processing and offers both Web UI and CLI interfaces.

Quick Start & Requirements

Installation is recommended via a portable package (ZIP) with a setup script, or manually via git clone and pip install -r requirements.txt. Requirements include Python 3.10+, PyTorch (auto-detected for various backends), font files (.ttf/.otf) in the fonts/ directory, and LLM/VLM access (API/local). Optional: Hugging Face token for OSB text, specific CUDA hardware for Nunchaku. First launch takes approximately 1-2 minutes.

Highlighted Details

• Supports 59 languages.
• Automated one-click translation.
• Handles speech bubble and non-speech bubble text.
• Offers Gradio Web UI and CLI.
• Includes 2x-AnimeSharpV4 upscaling.
• Supports custom font packs.

Maintenance & Community

The provided README does not contain specific details on notable contributors, sponsorships, community channels, or roadmaps.

Licensing & Compatibility

Licensed under Apache-2.0, permitting commercial use and integration into closed-source projects.

Limitations & Caveats

The Nunchaku backend (for FLUX.1 Kontext) is CUDA-only and requires NVIDIA 2000-series+ GPUs. OSB text pipeline setup needs a Hugging Face token for gated repos. Advanced model configurations may have specific hardware or setup prerequisites.