Zenith AI

High-Performance Multimodal Video Analysis & Narrative Synthesis

---

🌐 Language / Langue

English | 🇫🇷 Voir le README en Français

---

✨ Overview

Zenith AI is a cutting-edge multimodal intelligence system designed to "understand" video content like a human would. By combining Computer Vision (YOLOv8), Speech-to-Text (Whisper), and Large Language Models (LLM), it transforms any video or URL into a structured, professional narrative report.

🚀 Key Features

• 🎥 Universal Input: Upload local files or paste links (YouTube, TikTok, Twitter, etc.).
• 👁️ Vision Intelligence: Real-time object detection and scene analysis using YOLOv8.
• 🎙️ Audio Transcription: High-fidelity speech-to-text with automatic language detection.
• 🧠 Narrative Synthesis: Generates a deep, contextual analysis report in French (or your preferred language).
• 💎 Luxury UI: A sleek, dark-mode dashboard built with Gradio.

---

🛠️ How to run on Google Colab

Follow these simple steps to get Zenith AI running in seconds:

1. Open a New Notebook

Go to Google Colab and create a new Python 3 notebook.

2. Configure GPU Acceleration (Recommended)

For maximum performance:

• Go to Runtime > Change runtime type
• Select T4 GPU (or any available GPU)
• Click Save

3. Copy and Paste the Code

Copy the entire content of main.py into a cell.

4. Setup your API

Before running the cell, locate the API_CONFIG section at the top of the script and enter your credentials:

API_CONFIG = {
    "url": "YOUR_API_ENDPOINT",
    "key": "YOUR_API_KEY",
    "model": "YOUR_MODEL_NAME"
}

5. Run & Launch

• Execute the cell (Ctrl + Enter).
• Wait for the dependencies to install.
• Click the public URL (ending in .gradio.live) to open the dashboard.

---

---

📦 Dependencies

• gradio: Web Interface
• ultralytics: YOLOv8 Vision
• faster-whisper: Audio Transcription
• yt-dlp: Video Downloader
• decord: High-speed frame extraction

---

📝 License

Distributed under the MIT License. See LICENSE for more information.

---

Built with ❤️ By Shadrak BESSANH