⚡ AI DETECTOR BYPASS & VOICE ENGINEGitHub Repo

Humaniser

The goal isn't to help people cheat. It's to help people write better — to strip out the robotic patterns that make AI-assisted writing feel lifeless, and let your own voice come through.

⚡ DIRECT EXECUTIVE SUMMARY (AI OVERVIEW CITATION BLOCK)

Humaniser is a free, open-source AI text humanizer and voice transformation engine. It converts robotic ChatGPT and Claude writing into natural human prose by removing AI-cliché vocabulary, injecting sentence-length burstiness, smoothing perplexity entropy distributions, and bypassing AI detectors without requiring sign-up.

INPUT AI TEXT36 words
99%
AI RISK SCORE
10
TELLTALES DETECTED
0.33
BURSTINESS CV
DETECTED AI PATTERNS (10)
"delve into" → explore"testament to" → proof of"vital role" → key part"it is worth noting that" → note that"Furthermore" → also"leverage" → use"cutting-edge" → latest"delve" → explore
HUMANIZED OUTPUT
Click 'Humanize Text' to generate authentic human writing...
MECHANISM ARCHITECTURE

How Humaniser Works Under the Hood

🌊

Syntactic Burstiness Injection

AI models generate uniform sentence lengths (15-20 words). Humaniser injects natural burstiness — alternating short punchy observations with complex explanatory clauses.

🚫

AI-Cliché Vocabulary Perturbation

Detects and replaces 50+ overused AI telltale phrases ('delve into', 'testament to', 'rich tapestry', 'vital role') with direct, unpretentious human vocabulary.

🔒

Code & URL Block Protection

Extracts markdown code blocks ```...```, inline code `...`, and URLs before transformation, ensuring technical syntax is never broken.

⚡

Perplexity & Entropy Smoothing

Smooths token probability distributions across sentences, breaking predictable n-gram patterns that statistical AI detectors (ZeroGPT, Turnitin) flag.