What Happens When You Provide a Writing Sample for Voice Matching in Humanizer

When you provide a writing sample for voice matching, Humanizer analyzes your stylistic signals—including sentence rhythm, punctuation habits, and lexical preferences—to condition the rewrite, ensuring the output mirrors your personal voice rather than defaulting to a generic neutral tone.

The blader/humanizer repository provides a specialized text humanization skill that adapts its output to match an author's unique writing style. When a writing sample is provided for voice matching, the tool extracts specific linguistic features from your text and applies them to the final output, preserving distinctive quirks like em-dash usage or punchy sentence structures while still executing its core rewrite patterns.

How Voice Matching Works in Humanizer

The voice matching process operates through a three-phase pipeline that transforms generic AI output into prose that sounds authentically like the sample's author.

Sample Ingestion

The workflow begins when you paste 2–3 paragraphs of your own writing immediately after the prompt "Here's a sample of my writing for voice matching" within the /humanizer request. This raw text serves as the stylistic anchor for the entire transformation process.

Feature Extraction

Humanizer parses the provided sample to extract stylistic signals that define your voice. According to the repository's documentation in README.md (lines 50–64), the system specifically analyzes:

• Sentence length and rhythm – identifying whether you prefer crisp, short constructions or flowing, complex sentences
• Punctuation patterns – detecting distinctive habits such as frequent em-dash usage, semicolon deployment, or minimal comma usage
• Lexical preferences – noting your vocabulary choices, formality level, and transitional phrases
• Deliberate quirks – capturing unique elements like the strategic use of dashes for emphasis or specific parenthetical styles

Guided Rewrite

During the humanization phase, the model conditions its language generation on the extracted style while simultaneously applying the 25 standard rewrite patterns defined in SKILL.md. This dual process ensures that the core meaning remains intact, but the surface form adopts your rhythm, word choice, and punctuation habits. As implemented in the codebase, the final prose follows the sample’s specific cadence rather than the tool’s default neutral voice (lines 64–66, README.md).

Technical Implementation Details

The voice matching logic resides across two primary documentation files in the repository:

• SKILL.md – Contains the core skill prompt with all 25 rewrite patterns and the specific instructions for voice matching conditioning
• README.md – Houses the user-facing "Match your voice" section (lines 50–66) that explains the sample formatting requirements and expected behavior
• AGENTS.md – Provides guidance for AI agents on how to properly load and invoke the Humanizer skill with voice matching enabled

If no writing sample is provided, Humanizer defaults to its neutral voice, applying the same 25 rewrite patterns without the additional stylistic conditioning layer.

Code Example: Voice Matching in Practice

To activate voice matching, structure your request as follows:

/humanizer

Here's a sample of my writing for voice matching:
> I love crisp, short sentences—no fluff. My style is punchy, with lots of dashes—like this—because it feels alive.

Now humanize this text:
> The software automatically formats the output, making it look professional and readable.

Result (illustrative):

The software automatically formats the output—no fluff—making it look professional and readable, just like I prefer.

Contrast this with a standard request without voice matching:

/humanizer
Please humanize this text:
> AI tends to write long, flowing paragraphs that sound like a textbook.

[Paste your own sample above if you want a personal voice.]

Without the sample, the output defaults to a generic neutral rewrite without the distinctive dash-heavy, punchy style demonstrated in the first example.

Summary

• Providing a writing sample for voice matching triggers stylistic analysis of your sentence rhythm, punctuation habits, and word choice.
• The system extracts these signals from the sample section following the "Here's a sample of my writing for voice matching" prompt.
• SKILL.md contains the core rewrite logic, while README.md documents the voice matching user interface (lines 50–66).
• Output is conditioned on your personal style, preserving quirks like em-dash usage or short sentence structures.
• Without a sample, Humanizer defaults to a neutral voice that applies the 25 rewrite patterns without stylistic mirroring.

Frequently Asked Questions

How long should my writing sample be for effective voice matching?

You should provide 2–3 paragraphs of your own writing. This length provides sufficient data for the system to detect patterns in your rhythm, punctuation, and lexical preferences without overwhelming the context window or diluting the stylistic signals.

Does voice matching work with any type of content?

Yes, voice matching adapts to any writing genre you provide. Whether your sample is technical documentation, casual blog prose, or academic writing, the feature extracts the underlying stylistic markers—such as sentence length distribution and punctuation density—and applies them to the humanized output, regardless of the subject matter being rewritten.

Where is the voice matching logic defined in the codebase?

The voice matching instructions are primarily defined in SKILL.md, which contains the core skill prompt including the 25 rewrite patterns and the conditional logic for applying extracted stylistic features. User-facing documentation explaining how to format samples appears in README.md between lines 50 and 66, while AGENTS.md provides integration guidance for system agents loading the skill.

Can I use voice matching with the standard 25 rewrite patterns?

Yes, voice matching operates alongside the standard patterns rather than replacing them. The humanization process still applies the 25 rewrite patterns defined in SKILL.md, but it conditions the language generation on your extracted style. This means you receive the structural benefits of the rewrite patterns—such as varied sentence openings and reduced formulaic transitions—rendered in your personal voice rather than a generic neutral tone.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}