项目的 README,来自 github.com/AhmadHassan-BTed/Turnitout-Humanizer。合并为一个 markdown 文件:README.md。
The world's first 100% programmatic, zero-generative-AI footprint removal engine. Mathematically shatter cryptographic token trackers, purge invisible zero-width Unicode, and strip C2PA metadata. Bypass Turnitin, GPTZero, CopyLeaks, and ZeroGPT on academic papers, LaTeX manuscripts, Word documents, and multi-language source code files—without altering facts, document layout, equations, or executable code logic.
⚡ Looking for the Turnitout Enterprise REST API?
Integrate high-throughput, deterministic AI-evasion and smart citation generation directly into your SaaS platform, publishing pipeline, LMS, or backend infrastructure.
- 100% Zero-AI & Hallucination-Free: Pure mathematical pattern disruption, voice transformation, and syntax restructuring with zero LLM hallucination risk.
- Ultra-Low Latency: Asynchronous Python 3.14 + FastAPI + ASGI multi-worker engine (< 150ms single-paragraph latency, 200+ words/sec throughput).
- 100% Zero-Retention Privacy: Ephemeral in-memory execution. Uploaded text is never saved to disk, logged to databases, or used for AI training (GDPR & EU AI Act Article 50 compliant).
- Production Endpoints: High-speed string humanization (
POST /v1/humanize), multi-format document & ZIP archive in-place humanization (POST /v1/humanize/file), pre-flight file quote probe (POST /v1/humanize/file/quote), and contextual academic citation insertion with self-executing BibTeX generation (POST /v1/citations).📖 Read the Full Enterprise API Documentation & Schema Guide →
💬 Request API Keys & Cloud Access: Contact Ahmad Hassan directly on WhatsApp (+92 322 5522383) or email [email protected].
To comply with the EU AI Act (Article 50), OpenAI (ChatGPT), Anthropic (Claude), and Google (Gemini) now embed multi-layered tracking: invisible zero-width Unicode (crashing code/triggering detectors), hidden C2PA metadata, and cryptographic text watermarks like Google SynthID-Text. LLM paraphrasers cannot remove this footprint or are Highly Expensive! Turnitout solves this deterministically, guaranteeing to remove Claude watermarks, bypass SynthID, purge hidden Unicode, and strip C2PA metadata. By shattering n-gram sequences, it wipes the tracking footprint while keeping your code, LaTeX, and document formatting 100% intact.
| Cluster Mirror | Access URL | Status | Description |
|---|---|---|---|
| ⚡ Server Node 1 | turnitout1.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| 🚀 Server Node 2 | turnitout2.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| 🌟 Server Node 3 | turnitout3.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| 🔥 Server Node 4 | turnitout4.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| ✨ Server Node 5 | turnitout5.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| 💎 Server Node 6 | turnitout6.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| 🔮 Server Node 7 | turnitout7.streamlit.app | 🟢 Active | High-speed dedicated cluster mirror |
| ❤️ Main Server (Heart) | turnitout.streamlit.app | 🟢 Active | Resilient cluster anchor & active heart |
🔒 100% Privacy Guarantee: All documents and source code files are processed transiently in-memory. Zero document logging, zero server storage, zero data retention.
Commercial AI humanizers use LLMs to rewrite text, leaving new AI footprints, hallucinated facts, and cryptographic trackers intact.
Turnitout does NOT use AI. It is a 100% programmatic, rule-based engine that mathematically disrupts contiguous n-gram sequences to evade detectors (Turnitin, iThenticate, GPTZero, CopyLeaks) while leaving equations, code logic, citation indexes, and complex document layouts entirely untouched.
| Feature / Guarantee | 🤖 Commercial AI Humanizers | 🛡️ Turnitout Engine |
|---|---|---|
| New AI Footprint Risk | ❌ High (Generates new AI patterns) | ✅ 0% (100% Rule-Based) |
| Cryptographic Tracker Wipe (SynthID) | ❌ Fails (Preserves statistical tracking) | ✅ 100% (Shatters mathematical sequence) |
| Zero-Width Unicode Purge | ❌ Fails (Ignores hidden U+200B spaces) | ✅ 100% (Deep regex character scrubbing) |
| Fact & Data Preservation | ❌ Frequent Hallucinations | ✅ Zero (Facts & Numbers Stay Exact) |
| Document Layout Preservation | ❌ Destroys tables, XML & formatting | ✅ Preserves XML, Fonts & Headers |
| Source Code Safety | ❌ Breaks code syntax & logic | ✅ 100% Executable Logic Guarantee |
| LaTeX & Math Protection | ❌ Destroys \cite{}, equations |
✅ Keeps LaTeX & Equations Untouched |
| Reference Generation | ❌ None | ✅ BibTeX (.bib), LaTeX & JSON Generator |
| 🚨 Watermark & Tracker Wipe | 💻 Code Comment Humanizer | 📚 Citation & Reference Gen | 📁 Supported Formats |
|---|---|---|---|
| • Shatters Claude & SynthID • Purges Zero-Width Unicode • Strips C2PA Metadata |
• Python, C++, Java, JS, SQL • Developer Skill Personas • String-Aware Comment Safety |
• IEEE [1] & APA (Author)• BibTeX ( .bib) Generator• 5 Academic Taxonomies |
• .docx (Word Documents)• .tex (LaTeX Papers)• Source Code Files • Multi-file .zip Bundles |
.docx and file exports to ensure hidden Content Credential manifests are entirely wiped from the document history..docx, .tex, and .md files in-place. Your XML styling, fonts, tables, matrices, equations, and \cite{} tags are strictly protected."delve into", "pivotal role") and cleans em-dashes (—).#, //, /*, """) get rewritten..bib), LaTeX (\cite{key}), and Structured JSON reference packages..zip repo. It processes instantly, keeps your folder hierarchy intact, and spits out a 1-click ZIP export.| Architecture Layer | Core Component & Engine Responsibilities |
|---|---|
| 🖥️ UI Layer | Streamlit Multi-Page Web Application & Real-Time Dashboard |
| 🔑 Authentication Layer | OAuth 2.0 & Triple-Layer Auth Persistence Engine (Cookies + LocalStorage) |
| 📑 Parsing Layer | XML DOCX Parser, LaTeX Zone Parser & Multi-Language Code Tokenizer |
| ⚙️ Rule Transformation | 15-Stage Pipeline (Linguistic Synonyms, Active/Passive Voice, Clause Reshaping, N-Gram Audit) |
| 🗄️ Rules Repository | 4,068+ Decoupled Deterministic JSON Rules & Academic Discipline Taxonomy |
| 💻 Language Engines | String-Aware Token Matching for Python, C-Family, Shell, SQL & HTML |
| 🔍 Verification Layer | ZipPy Entropic Compression Perplexity & Real-Time Evasion Scorer |
The platform is designed around strict asset decoupling and modular language rule pipelines:
.docx (XML styling, bold, italics, tables, and headers preserved in-place).xlsx, .csv (Table cells and text columns humanized without breaking formulas).tex (Math environments $ ... $, \[ ... \], equation blocks, and citations strictly preserved).txt, .md.py, .cpp, .c, .java, .cs, .js, .ts, .go, .rs, .sh, .sql, .html, .css (100% executable code preservation).zip (Full multi-file repository hierarchies extracted, humanized, and packaged back into 1-click ZIP export)If Turnitout has saved you time, protected your research, or helped your stack, consider supporting ongoing development and server infrastructure costs!
Because GitHub Sponsors and Stripe do not currently support Pakistan, you can send support globally in under 60 seconds:
Atta9196 (Account: Ahmad Hassan)United Bank Limited (UBL Ameen) • IBAN: PK07UNIL0109000387828585 • BIC: UNILPKKA03225522383 / PK44SADA0000003225522383) or UBL Ameen (2523387828585)📖 Read the Full Support Guide with Binance Pay QR Code →
Built with ❤️ by Ahmad Hassan for researchers, developers, students, and academics worldwide.