YANG-Haruka/LinguaHaru

Next-Gen AI Translation Tool Powered by LLM. Support Office documents, PDF, TXT, and more format with just one click.

What it solves

LinguaHaru provides a unified, high-quality AI translation solution that handles complex file formats, media, and live speech without requiring technical installation. It solves the problem of fragmented translation tools by offering a single, crash-safe engine that preserves document layouts and formatting while supporting everything from office documents to scanned manga and real-time audio.

How it works

The project uses a modular architecture where a core translation engine handles batching, caching, and glossary enforcement, while specialized plugins manage specific tasks like OCR, PDF layout preservation, or speech-to-text. Users can choose between a native desktop app or a web console, both of which connect to AI APIs (like DeepSeek) or local models. The system automatically detects hardware to enable GPU acceleration for heavy tasks and handles all dependency management internally.

Who it’s for

It is designed for anyone needing professional-grade, automated translation for documents, media, or live conversations. It is particularly useful for users who want a "set-it-and-forget-it" experience, as the portable builds require no Python or CUDA configuration, and for those who need to process large batches of files reliably with the ability to resume interrupted jobs.

Highlights

  • Comprehensive Format Support: Handles Office documents (DOCX, PPTX, PDF), subtitles, images, and video/audio files.
  • Crash-Safe Reliability: Features atomic incremental writes, allowing jobs to resume exactly where they left off if interrupted.
  • Plugin-Based Architecture: Keeps the base installation small; users only install the specific plugins (e.g., Manga mode, Real-Time Voice) they need.
  • Smart Engine: Includes glossary enforcement, placeholder protection, and post-translation quality checks to ensure accuracy.
  • Flexible Deployment: Offers a native desktop app (Windows/macOS), a web console for LAN/multi-user access, and Docker support.
  • Regional Optimization: Automatically uses fast mirrors for dependencies and models, making it accessible in regions with restricted network access without needing a VPN.

Related

  • Project
    RManLuo/gfm-rag

    GFM‑RAG is a Python library that builds a graph index from documents (or an existing knowledge graph) and uses a pre‑trained Graph Foundation Model—a graph neural network—to retrieve relevant passages for Retrieval‑Augmented Generation. It supports out‑of‑the‑box use on new datasets, fine‑tuning for specific domains, and provides interpretable multi‑hop reasoning paths.

  • Project
    superzhang21/ghostwriter

    Ghostwriter is a library of deeply-modelled AI character personas (like Lu Xun, Ding Yuanying, etc.) with a zero-dependency Python script that generates ready-to-paste prompts, letting any LLM mimic a chosen persona's speech, thinking and vocabulary.

  • Project
    stanford-oval/suql

    A query language that extends SQL with free-text primitives to enable conversational search over databases containing both structured and unstructured data.

  • Project
    yolo-hyl/medical-rag

    A professional medical RAG system using LangChain and Milvus that employs hybrid retrieval and intelligent agents to provide high-accuracy medical question-answering.