---

🚀 VeriSight — Agentic AI Fake Video Detection

A prototype full-stack system for deepfake detection, evidence packaging, and cryptographic verification. This monorepo contains a FastAPI backend and a React + Vite frontend, integrated as an autonomous analysis and review pipeline.

📚 Table of Contents

• Overview
• Badges
• Logo
• Quick TL;DR
• Key Features
• Installation
• Run Locally
• Environment Variables
• Configuration & Proxy
• Architecture & Tech Stack
• Usage / Examples
• API Reference
• How It Works (Pipeline)
• Preprocessing & Detector
• Verification Pack & Cryptography
• Evidence Viewers
• Testing
• Deployment
• Roadmap
• Contributing
• License
• Authors
• Acknowledgements
• FAQ
• Support / Contact
• Related
• Troubleshooting
• Changelog
• Optimizations / Lessons Learned
• Used By / Integrations
• Custom Sections

---

🧾 Overview

VeriSight is an autonomous AI pipeline designed to detect manipulated or AI-generated videos and generate tamper-proof verification packs signed cryptographically. Frontend UI reference: verisight-frontend/src/pages/public/Home.jsx

---

🏷 Badges

Build	License	Release
(CI Badge placeholder)	Proprietary	PyPI/NPM (if published)

---

🔰 Logo

[logo](https://github.com/user-attachments/assets/7fb2a0a7-2eac-460c-8fec-ec23138be1a8)

---

⚡ Quick TL;DR

Component	Path
FastAPI backend	backend/app.py & backend/app2.py
React + Vite frontend	/verisight-frontend
Local storage mount	/storage in backend/app.py
Detector core	analyze_video() in backend/src/ai_video_detector.py

---

✨ Key Features

• Multimodal video analysis (visual, audio, temporal)
• Frame extraction & optical-flow-based analytics
• SSIM/MSE anomaly scoring + robust z-score voting
• Cryptographically signed verification pack
• Evidence review UI: frames, heatmaps, spectrograms
• Upload via file or URL with progress tracking

---

🛠 Installation

Backend

See backend/README.md

sudo apt update && sudo apt install -y python3-pip python3-venv ffmpeg
python3 -m venv .venv
source .venv/bin/activate
pip install -r backend/requirements.txt
uvicorn backend.app:app --reload --port 8000

Frontend

npm install
npm run dev

---

▶ Run Locally

Backend

uvicorn backend.app:app --reload --port 8000

Artifacts available via /storage

Frontend

npm run dev

Requires VITE_API_BASE_URL configured in .env.example

---

🔐 Environment Variables

Key	Location
VITE_API_BASE_URL	verisight-frontend/.env.example
Backend config	backend/config.py

---

🏗 Architecture & Tech Stack

   /-----------------------\
  /  UI (React + Vite)     \
 /-------------------------\
|  API Layer (FastAPI)     |
| Routers: auth, upload,   |
| jobs, analysis           |
 \-------------------------/
  \ Worker / FFmpeg / Proc /
   \-----------------------/

Processing Stack

System	Tooling
Preprocessing	FFmpeg (extract_frames, make_preview_clip)
Detector	analyze_video, optical flow, SSIM/MSE
Storage	SQLite + local filesystem

---

🧪 Usage / Examples

Upload via CLI

curl -F "file=@/path/to/video.mp4" http://127.0.0.1:8000/upload

---

🔑 API Reference

Area	File
Upload	backend/src/routes_upload.py
Analysis	backend/src/routes_analysis.py
Jobs	backend/src/routes_jobs.py
Auth	backend/src/routes_auth.py

---

🧠 How It Works — Pipeline

1. Upload video (file/URL)
2. FFmpeg preprocessing (utils_ffmpeg.py)
3. Multimetric detector (ai_video_detector.py)
4. Evidence generation & previews
5. JSON/PDF cryptographic verification pack
6. Human review UI

---

🧾 Verification Pack & Cryptography

Frontend UI: VerificationPack.jsx Signature block component: SignatureBlock.jsx

---

🔎 Evidence Viewers

Viewer	File
Frames	FrameGrid.jsx
Heatmaps	HeatmapViewer.jsx
Spectrograms	SpectrogramPlayer.jsx

---

📦 Deployment

• npm run build for frontend
• Serve backend with gunicorn + uvicorn
• Ensure ffmpeg installed on host

---

🛣 Roadmap

• ML-based detector replacing heuristics
• Celery/RQ workers + S3 storage
• Blockchain anchor for audit logs
• Enterprise SSO & roles

---

💼 Contributing

• ESLint + Prettier + Black/flake8 recommended
• Add PR templates and CI testing

---

📜 License

Proprietary — modify as required

---

👥 Authors

Add contributors list here

---

🙏 Acknowledgements

FFmpeg • OpenCV • scikit-image • FastAPI • React • Vite

---

❓ FAQ

Question	Answer
Supported formats?	MP4, MOV, AVI, WebM
Where are files stored?	Local UPLOAD_DIR (config.py)

---

🆘 Support / Contact

privacy@verisight.ai (placeholder)

---

🔧 Troubleshooting

• FFmpeg errors → ensure installed & in PATH
• Preview fails → review utils_ffmpeg.py

---

🧠 Optimizations / Lessons Learned

• Downscaled grayscale frames improve temporal anomaly precision
• Fusion of MSE/SSIM/flow metrics yields stable detection

---

🧩 Used By / Integrations

• Coming Soon (Cloudflare / Stripe etc.)

---

📌 Custom Sections

Add brand-specific or hackathon-specific notes here

---

🌟 Done!