Practical overview
When it fits
Use it when you have image-based text to translate and can work with a local Python application or set up its documented web or API mode.
Who it is for
- Readers who want to understand text in manga or other images, particularly Japanese-language material.
- Developers seeking a Python project for image-text detection, translation, repair, and typesetting.
Documented uses
- Translate text embedded in a manga page or other image.
- Process a local image or folder in batch mode.
- Remove original lettering and typeset translated text.
Requirements & limits
The project is publicly available as source code under the GPL-3.0 license. The README marks its older online-service address as no longer working.
- The README lists Japanese as the main supported language, alongside Simplified and Traditional Chinese, English, and 20 other minor languages.
- Local installation requires Python 3.12 or later and may require Git and a C++ compiler for dependencies.
- GPU use requires a PyTorch installation matched to the CUDA version.
- The project cautions that it is early-stage and that small or irregular text can challenge detection.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up a local Python installation
Install the project dependencies for local use.
Before you start
- Python 3.12 or later
- Git and a C++ compiler for building a dependency
- On Windows, Microsoft C++ Build Tools
Steps
Clone the project repository.
Create and activate a Python virtual environment if desired.
Install a PyTorch build that matches the CUDA version if GPU use is intended.
Install project dependencies with pip from the requirements file.
Expected outcome
Project dependencies are installed; models are downloaded to the ./models directory at runtime.
Guide limitations
- The README cautions that the very latest Python version may not work with some PyTorch libraries.
- GPU setup requires a PyTorch version matching the CUDA installation.
- This setup guide does not cover running a translation.
Translate a local image or image folder
Run the documented local batch workflow on a file or folder.
Before you start
- A local installation of the project and its dependencies
- A path to an image file or image folder
Steps
Provide the image file or folder path to the local-mode command.
Run the command using the documented local mode and input option.
Check the adjacent translated-output folder named with the -translated suffix.
Expected outcome
The translated results are placed in a folder based on the input folder path.
Guide limitations
- The project describes itself as early-stage and notes it has shortcomings.
- Small images and irregularly sized text can make detection more difficult.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- zyddnys/manga-image-translator · Official homepage title · retained in its source language