koharu — What is it?

Koharu is an open-source, ML-powered manga translator that automates the process of translating manga pages while preserving the original layout and style.

⭐ 2,371 Stars 🍴 124 Forks Rust — Author: mayocream
Source: Official README and linked documentation View on GitHub →

Why it matters

Koharu is being discussed because its innovative approach to manga translation, addressing the need for accurate and contextually appropriate translations while maintaining the visual integrity of the original manga. Its use of local-first workflow and advanced ML models for OCR, inpainting, and translation differs through the open-source space.

Source: Official README and linked documentation

Core Features

Automatic Text Detection and Cleanup

Koharu automatically detects text regions, speech bubbles, and cleanup masks, using object detection and OCR to ensure accurate translation without altering the original page layout.

Source: Official README and linked documentation
Inpainting and Translation

The project employs inpainting models to remove source lettering and uses local or remote LLM backends for translation, ensuring privacy and flexibility in the translation process.

Source: Official README and linked documentation
Advanced Text Rendering

Koharu supports vertical CJK and RTL text rendering, with font fallback and vertical punctuation alignment, to ensure translated text reads naturally within the manga context.

Source: Official README and linked documentation
Layered PSD Export

The tool allows for exporting translated pages as layered PSD files, preserving helper layers and text layers for further editing and refinement.

Source: Official README and linked documentation
Local HTTP API and MCP Server

Koharu includes a local HTTP API and MCP server for automation and integration with other tools, enhancing its versatility for various use cases.

Source: Official README and linked documentation

Architecture

Koharu's architecture is modular, with separate components for vision and language processing. It uses Rust for performance and safety, use Tauri for the desktop app and candle and llama.cpp for ML inference. The project employs a staged stack of models for different stages of the translation process, ensuring efficient and accurate translations.

Source: Official repository tree and dependency files

Project Knowledge Graph

Knowledge graph: project (center) + core features (inner hexagons) + key dependencies (outer chips) candle-transformerscandle-transfo… candle-core candle-nn candle-flash-attncandle-flash-a… hf-hub Automatic Text Detection and CleanupAutomatic Text Dete… Inpainting and TranslationInpainting and Tran… Advanced Text RenderingAdvanced Text Rende… Layered PSD Export Local HTTP API and MCP ServerLocal HTTP API and… koharu Project Core feature Key dependency

Tech Stack

LanguageRustFrameworkTauri, candle, llama.cpp
candle-transformerscandle-corecandle-nncandle-flash-attnhf-hubimageanyhowasync-traitimageprocclapserdeserde_bytesserde_jsontokiotracingtracing-subscriberconsoleopenrfdpostcardthiserrortauritauri-build
Not enough information.
Source: Official repository tree and dependency files

Quick Start

To install Koharu, run the following command: ```bash koharu --port 9999 ```
Source: Official README and linked documentation

Use Cases

Koharu is suitable for manga enthusiasts, translators, and developers looking to automate the translation of manga pages. It is useful for creating localized versions of manga, facilitating accessibility, and enabling new audiences to enjoy manga content.

Source: Official README and linked documentation

Strengths & Limitations

Strengths

  • High accuracy and contextually appropriate translations
  • Preserves original manga layout and style
  • Local-first workflow for privacy
  • Modular architecture for scalability

Limitations

  • Limited documentation and community support
  • May require technical expertise to set up and use
  • Some features may be experimental or not fully developed
Source: Official README and linked documentation

Latest Release

As of 2026-09-23, the latest stable Release is 0.83.5, published on 2026-09-22. It mainly fixes incomplete Release-asset uploads and provides 22 cross-platform assets; 0.83.0 contains the larger feature update.

Source: github.com/koharu-rs/koharu/releases/tag/0.83.5

Verification and evidence

status: docs-only  |  checked_at: 2026-09-23  |  environment: Only the migrated official repository, official documentation, hardware notes, and GitHub Releases were reviewed.  |  limitations: No local installation or runtime verification was performed. No package, runtime, or model was downloaded, no.khrproj was created, and OCR, translation, inpainting, GPU acceleration, and PNG/PSD export were not tested.  |  observed_output: The latest official Release is 0.83.5, published on 2026-09-22, with Windows, macOS, and Linux assets. It repairs incomplete Release-asset uploads; 0.83.0 added CBZ export, page-range prompts, and local translation-model download status.

Source: github.com/koharu-rs/koharu/releases/tag/0.83.5

Comparison

The official manga-image-translator project centers on a Python CLI, batch processing, Web/API servers, and extensive configuration for scripted directory translation or service calls. Koharu emphasizes desktop projects, a visual canvas, layer-by-layer review, and layered PSD export. Choose the former for command-line pipelines and Koharu for interactive repair and typesetting.

Source: github.com/zyddnys/manga-image-translator

Verdict

Koharu is a open-source project for manga translation, offering a unique combination of advanced ML capabilities and a user-friendly interface. It is particularly suitable for individuals and teams interested in manga localization and accessibility, though it may require some technical knowledge to fully utilize its features.

Frequently Asked Questions

What did the documentation review verify for koharu on 2026-09-23?

The latest official Release is 0.83.5, published on 2026-09-22, with Windows, macOS, and Linux assets. It repairs incomplete Release-asset uploads; 0.83.0 added CBZ export, page-range prompts, and local translation-model download status. No local installation or runtime verification was performed.

What evidence boundary applies to this koharu review?

The review environment was: Only the migrated official repository, official documentation, hardware notes, and GitHub Releases were reviewed. koharu behavior outside that scope still needs a local check.

Sources and review scope
This review is based on the project's public documentation. Versions, installation commands, and feature boundaries can change upstream; check the cited sources and current release before use. Sources checked on: 2026-09-23 13:55.

Reference materials: README, GitHub API, dependency files