🎥 InsightFlow AI

Transform YouTube videos into searchable knowledge with AI-powered transcription and semantic search

InsightFlow AI is an intelligent video processing system that extracts audio from YouTube videos, transcribes content using OpenAI's Whisper, and creates a searchable question-answering system using LangChain's RAG (Retrieval-Augmented Generation) capabilities.

✨ Features

• 🎯 AI-Powered Transcription: Automatically transcribe YouTube videos using OpenAI's Whisper model
• 🧠 Semantic Search: Query video content using natural language questions
• 💬 Interactive Q&A: Ask specific questions and get relevant answers from the video content
• 📊 Full Transcript Access: View and download complete transcriptions
• 🔍 Vector Database: Leverages ChromaDB for efficient semantic search
• 🚀 Modern UI: Clean, responsive interface built with Streamlit

🛠️ Tech Stack

• Frontend: Streamlit
• Transcription: OpenAI Whisper
• Vector Database: ChromaDB
• Embeddings: Sentence Transformers (all-MiniLM-L6-v2)
• LLM Framework: LangChain
• Video Processing: yt-dlp, FFmpeg

📋 Prerequisites

Before you begin, ensure you have the following installed:

• Python 3.8 or higher
• FFmpeg (Download here)
• pip (Python package manager)

🚀 Installation

1. Clone the Repository

git clone https://github.com/yourusername/insightflow-ai.git
cd insightflow-ai

2. Create Virtual Environment

# Windows
python -m venv venv
venv\Scripts\activate

# macOS/Linux
python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Install FFmpeg

Windows:

1. Download FFmpeg from ffmpeg.org
2. Extract to a directory (e.g., D:\ffmpeg\bin)
3. Add to System PATH or update the path in processor.py

macOS:

brew install ffmpeg

Linux:

sudo apt update
sudo apt install ffmpeg

💻 Usage

Running the Application

streamlit run app.py

The application will open in your default browser at http://localhost:8501

Using InsightFlow AI

1. Paste YouTube URL: Enter the URL of the YouTube video you want to analyze
2. Click "Analyze Video": Wait for the processing to complete (1-3 minutes depending on video length)
3. Ask Questions: Once processing is complete, ask questions about the video content
4. View Transcript: Expand the transcript section to see the full text

Example Questions

• "What is the main topic of this video?"
• "Can you summarize the key points discussed?"
• "What does the speaker say about [specific topic]?"
• "What are the recommendations mentioned?"

📁 Project Structure

insightflow-ai/
│
├── app.py                 # Main Streamlit application
├── processor.py           # Video download and transcription logic
├── brain.py              # Vector database and RAG implementation
├── requirements.txt      # Python dependencies
├── README.md            # Project documentation
│
├── .vscode/
│   └── launch.json      # VS Code debug configuration
│
├── venv/                # Virtual environment (not tracked)
├── chroma_db/           # Vector database storage (generated)
└── temp_audio.mp3       # Temporary audio files (generated)

🔧 Configuration

FFmpeg Path (Windows Users)

If FFmpeg is not in your system PATH, update the path in processor.py:

os.environ["PATH"] += os.pathsep + r"YOUR_FFMPEG_PATH\bin"

Whisper Model Selection

You can change the Whisper model for different accuracy/speed tradeoffs in processor.py:

# Options: tiny, base, small, medium, large
model = whisper.load_model("base")  # Change "base" to your preferred model

Model	Speed	Accuracy	Use Case
tiny	⚡⚡⚡	⭐⭐	Quick testing
base	⚡⚡	⭐⭐⭐	Default, balanced
small	⚡	⭐⭐⭐⭐	Better accuracy
medium	🐌	⭐⭐⭐⭐⭐	High accuracy
large	🐌🐌	⭐⭐⭐⭐⭐	Best accuracy

🧪 Development

Running in Debug Mode (VS Code)

1. Open app.py in VS Code
2. Press F5 or click "Run and Debug"
3. Select "Python: Streamlit" configuration

Testing Individual Components

Test Processor:

python processor.py

Test Brain (Vector DB):

python brain.py

📦 Dependencies

yt-dlp                    # YouTube video downloader
openai-whisper           # Audio transcription
langchain-text-splitters # Text chunking
langchain-community      # LangChain integrations
langchain-core           # LangChain core functionality
chromadb                 # Vector database
sentence-transformers    # Text embeddings
torch                    # PyTorch for ML models
streamlit                # Web interface

🎯 Use Cases

• 📚 Educational Content: Extract key information from lectures and tutorials
• 🎙️ Podcast Analysis: Search through podcast episodes for specific topics
• 📺 Video Research: Quickly find relevant sections in long-form content
• 📝 Meeting Recordings: Create searchable transcripts of recorded meetings
• 🎬 Content Creation: Analyze competitor videos or research topics

🛣️ Roadmap

• Support for multiple video sources (Vimeo, local files)
• Multi-language support
• Export functionality (PDF, DOCX)
• Timestamp-based search results
• Video player integration with auto-jump to relevant sections
• Batch processing for multiple videos
• Advanced analytics dashboard

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• OpenAI Whisper for the incredible speech recognition model
• LangChain for the RAG framework
• Streamlit for the easy-to-use web framework
• ChromaDB for the vector database

👤 Author

Minhajul Islam Nion

• Email: minhajulislamnion@gmail.com
• University: University of Canberra
• LinkedIn: Nion007
• GitHub: @Nion9

📧 Contact

For questions or feedback, please reach out via email or open an issue on GitHub.

---

Built for recruiters and AI enthusiasts