KeywordGacha (KG) is a next-generation translation assistance tool designed to streamline the creation of terminology lists for translators working with novels, games, and subtitles. It leverages AI to automatically analyze Chinese, English, and Japanese text, generating entity term lists, translating terms, summarizing character information, and identifying key entities to ensure consistent translation of names and terms across long-form content.

How It Works

KG utilizes AI large language models (LLMs) compatible with the OpenAI API standard. Users can opt for local LLM inference, requiring at least 8GB of VRAM on an Nvidia GPU, or cloud-based services like Volcano Engine or DeepSeek for faster, higher-quality results without specific hardware demands. The tool processes various text formats, extracts entities, and generates detailed logs and term lists in JSON and XLSX formats.

Quick Start & Requirements

• Install: Download application files (.001, .002) from the release page or Baidu Netdisk, place them in the same folder, and decompress .001.
• Configuration: Edit config.json to input API keys and base URLs for your chosen AI model provider (local or cloud).
• Run: Execute app.exe. Processed results are saved in the output folder.
• Prerequisites: OpenAI-compatible LLM API access (local or cloud), Python (implied by .exe but not explicitly stated).
• Resources: Local inference requires an Nvidia GPU with >= 8GB VRAM. Cloud options have no specific hardware requirements.
• Links: Release Page (implied), Baidu Netdisk (implied), Discord: https://discord.gg/pyMRBGse75

Highlighted Details

• Achieves 80%-90% hit rate compared to manually created term lists on test samples.
• Supports various text formats including .srt, .ass, .txt, .epub, .md, .rpy, .json, .xlsx, and .trans.
• Features include automatic translation, summarization, and analysis of character information.
• Recent updates add support for RTX 50 series GPU acceleration and WOLF official translation tool exports.

Maintenance & Community

The project is actively maintained with recent updates in April 2025. Community support is available via QQ (417632316) and Discord.

Licensing & Compatibility

The README states: "如您的项目涉及任何商业行为或者商业收益，在使用 KeywordGacha 前，请先与作者联系以获得授权！" (If your project involves any commercial activities or commercial gains, please contact the author for authorization before using KeywordGacha!). This implies a non-commercial or restrictive license for commercial use.

Limitations & Caveats

The tool has a maximum single-line processing limit of approximately 500 characters; longer lines may be truncated. For optimal results with novel text, unrelated content like author comments should be pre-removed. Russian language support is planned but not yet implemented.