devswha/patina

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

What it solves

patina addresses the problem of "AI-sounding" text—phrasing that feels robotic or formulaic—by rewriting it to sound more natural and human. Unlike tools designed to evade AI detectors, it focuses on maintaining the original meaning, facts, and intent while stripping away typical AI linguistic patterns.

How it works

The tool uses a deterministic, pattern-based approach across four languages (English, Korean, Chinese, and Japanese). It employs a catalog of 184 named patterns to identify and rewrite AI-typical phrasing. To ensure the original message isn't lost, every rewrite is gated by meaning-preservation (MPS) and fidelity floors; if a rewrite drifts too far from the original meaning, it is retried or rolled back. Users can control the output using three independent axes: Document Type (genre/purpose), Persona (voice fingerprint), and Register (casual vs. professional).

Who it’s for

It is designed for writers and developers who use AI for drafting but want the final output to feel human-authored. It is available as a Node CLI, a browser-based playground, and as a skill for AI agents like Claude Code, Cursor, and Codex.

Highlights

  • Meaning Preservation: Uses MPS and fidelity checks to ensure claims, numbers, and causation remain unchanged.
  • Auditable Process: Not a black box; users can use --diff to see exactly which patterns triggered a change.
  • Multi-Axis Control: Independent settings for document genre, personal voice, and delivery register.
  • Broad Integration: Works via CLI, browser, or directly within AI agent environments.

Related

  • Project
    blader/humanizer

    A tool that rewrites AI-generated text to sound more human by applying 35 patterns based on Wikipedia's signs of AI writing, while preserving original facts and meaning.

  • Project
    LifelongLazyLearner/qu-ai-wei

    A Simplified Chinese rewriting skill that removes stiff, formulaic AI-generated patterns to make text sound more natural while preserving original facts and tone.

  • Project
    petergyang/no-ai-slop

    A tool that removes over 20 common patterns of generic AI-generated writing to help authors maintain a personal, human voice in their AI-assisted drafts.

  • Project
    epoko77-ai/im-not-ai

    A Claude Code and Codex CLI skill that removes AI-typical patterns from Korean text using a research-based taxonomy to make it sound natural without altering the original meaning.

  • Project
    KKKKhazix/human-writing

    A framework and toolset for AI agents to eliminate "AI-style" Chinese writing by enforcing human-like rhythms, factual substance, and structural checks.