neavo/LinguaGacha

LinguaGacha – AI‑powered batch translator for novels, games and subtitles

What it does

• A desktop application that lets you drop a file (subtitle, ebook, game script, etc.) and get a full‑text translation in seconds.
• Supports 16 languages (Chinese, English, Japanese, Korean, Russian, German, French, Italian …) and a variety of file formats such as .srt, .ass, .epub, .rpy, .json, .xlsx.
• Uses large‑language‑model APIs (OpenAI, Anthropic, Google, DeepSeek, local models like Qwen2.5‑7B, SakuraLLM) behind an “AGENT” mode that runs a scripted conversation: term extraction → bulk translation → optional automatic proofreading.

Why it matters

• Translating visual‑novel‑style games or long novels is usually a manual, error‑prone process. LinguaGacha automates the pipeline while preserving formatting, style tags and code‑like markup, which saves hours of post‑editing.
• The built‑in terminology table keeps character names and special terms consistent across an entire work, a common pain point for fan‑translation projects.

How to get started

1. Download the appropriate binary from the Releases page (Windows .zip, macOS .dmg, Linux .AppImage).
2. Install an LLM endpoint – the README recommends DeepSeek because it works without a GPU.
3. Drag‑and‑drop the source file onto the app, set source/target languages, pick a model in the AGENT panel, and click the preset commands to run term extraction, translation and optional proofreading.
4. The translated file is saved with the original formatting intact.

Key technical bits

• Front‑end built with Electron/Node.js (the repo mentions npm run dev and npm run build).
• The translation logic is a thin wrapper around LLM chat completions; the “AGENT” mode strings together several prompts to achieve term extraction, translation, and quality‑check.
• Supports both cloud APIs and local inference models, so users with a capable GPU can run models like Qwen2.5‑7B locally.

Who it’s for

• Hobbyists and fan‑translation groups who need to localise visual‑novel, RPGMaker, RenPy, WOLF, or other engine games.
• Readers who want to quickly translate ebooks or subtitles without fiddling with command‑line tools.
• Anyone looking for a one‑click, GUI‑driven way to harness LLM translation power.

Community & support

• Official wiki with tutorials, best‑practice guides for specific game engines, and a CLI mode for power users.
• Discord and QQ groups for real‑time help, plus an issue tracker for feature requests.

License & attribution

• Distributed under an open‑source license (badge shown in the README). Users must credit LinguaGacha in any public work that uses its translations, and commercial use requires author permission.

---

All information above is taken directly from the project's README.