devswha/patina

AI-writing humanizer for KO/EN/ZH/JA

What it solves

Patina is designed to remove "AI-sounding" phrasing from text in Korean, English, Chinese, and Japanese. It focuses on humanizing AI-assisted drafts while ensuring that the original meaning, numbers, polarity, and causation remain unchanged, rather than acting as a tool to evade AI detectors.

How it works

The tool uses a deterministic, pattern-based approach with 189 named patterns to identify and rewrite AI-typical phrasing. It employs "meaning preservation" (MPS/fidelity) gates to verify that rewrites do not drift from the original claim. Users can customize the output through three independent axes: Document Type (genre/purpose), Persona (voice fingerprint), and Register (casual or professional delivery).

Who it’s for

It is intended for writers and developers who use AI for drafting and want their final text to sound more natural and human, as well as those integrating humanization tools into agentic workflows (e.g., via Claude Code, Cursor, or Node CLI).

Highlights

  • Auditable Rewrites: Uses named patterns and a --diff mode to show exactly why and how text was changed.
  • Multilingual Support: Works across English, Korean, Chinese, and Japanese.
  • Meaning Verification: Includes a --verify mode to check rewrites against fidelity floors to prevent meaning drift.
  • Flexible Integration: Available as a browser playground, Node CLI, and as a skill for various AI agents.
  • Voice Learning: Ability to learn a reusable voice fingerprint from a provided sample using the persona command.

Related

  • Project
    epoko77-ai/im-not-ai

    A CLI tool for Claude Code, Copilot, and Codex that removes AI-specific patterns and translationese from Korean text to make it sound natural and human-like without changing the original meaning.

  • Project
    blader/humanizer

    Humanizer is a plug‑in for Claude Code, Codex and other LLM‑powered agents that rewrites AI‑generated text to sound like a real person, using a rule‑based detector for 26 common “AI‑writing tells”. It installs via the Skills CLI, is invoked with `/humanizer`, can match a user’s voice, and preserves factual content without hallucination.

  • Project
    conorbronsdon/avoid-ai-writing

    A portable writing skill for AI agents that audits and rewrites content to remove formulaic AI writing patterns and "AI-isms" using a tiered vocabulary and pattern catalog.

  • Project
    theclaymethod/unslop

    A tool that detects and removes formulaic AI writing patterns while preserving meaning and voice, featuring a three-layer detection system and personalized voice mimicry.

  • Project
    larashero3-dotcom/lieflat-less-ai-tone

    A linguistic research project and toolset that quantifies "AI flavor" in Chinese text and provides a set of rules to remove these markers for more human-like writing.