Summary

This project provides an open-source pipeline to "humanize" AI-generated text, aiming to bypass AI detection systems like Turnitin and GPTZero. It targets users needing to make AI content sound more natural and less machine-generated across multiple languages, offering a free and adaptable solution.

How It Works

The core approach employs a multi-stage LLM and translation rewriting process. Text is first rewritten by DeepSeek LLM, translated to Chinese for structural variation, then to Turkish via Google Translate for syntactic distortion. An optional DeepL translation step (Turkish to Japanese) further enhances linguistic diversity. Finally, DeepSeek rewrites the text back to the original language, preserving semantics while reconstructing natural flow. This pipeline leverages cross-lingual transformations to break AI patterns.

Quick Start & Requirements

Installation involves cloning the repository, installing dependencies via pip install -r requirements.txt, and copying a configuration file (config.example.toml to config.toml). The primary execution command is python main.py. Key requirements include a DeepSeek API key for core functionality. An optional DeepL API key can be provided for an additional translation stage. Configuration options in config.toml allow setting the target language, API keys, LLM model parameters (temperature, slug), and base URLs.

Highlighted Details

• Effectively bypasses most AI detection tools (e.g., Turnitin, GPTZero).
• Supports 8 languages: English, Japanese, Chinese, Korean, German, French, and Spanish.
• Fully free and open-source, with an MIT license.
• Preserves original semantic meaning while enhancing naturalness and stylistic diversity.

Maintenance & Community

The provided README does not detail specific contributors, community channels (e.g., Discord, Slack), or a public roadmap.

Licensing & Compatibility

The project is released under the MIT License, which is highly permissive and generally compatible with commercial use and closed-source linking.

Limitations & Caveats

Core functionality necessitates a DeepSeek API key. While an optional DeepL API key enhances diversity, it adds complexity and potential cost. The effectiveness against all AI detectors is not guaranteed. The pipeline's reliance on multiple external translation services introduces potential points of failure, latency, and subtle semantic drift.