Transform YouTube comment sections into structured insights using AI-powered analysis. Extract meaningful patterns, sentiment, and answers from thousands of comments automatically.

• Comment Extraction: Fetches all comments and replies using YouTube Data API
• AI Summarization: Generate comprehensive summaries of discussion themes
• Sentiment Analysis: Classify comments as positive, negative, or neutral
• Visual Analytics: Sentiment charts and word clouds
• Vector Search: Semantic search through comment databases

• Natural Language Queries: Ask questions about the video comments
• Context-Aware Responses: Answers based on actual comment content
• Adaptive Sampling: Intelligently selects relevant comments for accuracy
• Multi-video Support: Maintains separate databases per video

• Small videos (< 500 comments): 2-3 minutes
• Medium videos (500-2000 comments): 3-5 minutes
• Large videos (2000+ comments): 5-10+ minutes

• Comment retrieval: Milliseconds (vector search)
• Answer generation: 30-60 seconds (LLM processing)
• Total Q&A time: ~1 minute per question

Performance depends heavily on hardware specifications (GPU vs CPU).

• Python 3.8+
• Ollama installed and running
• YouTube Data API key (Get one here)
• Hardware Recommendations:
  • 8GB+ RAM (16GB for large comment sets)
  • GPU with CUDA support (optional but significantly faster)
  • 5GB+ free storage for models and vector databases

# Clone the repository
git clone https://github.com/yourusername/youtube-comment-summarizer
cd youtube-comment-summarizer

# Install dependencies
pip install -r requirements.txt

# Download required models
ollama pull llama3.2
ollama pull mxbai-embed-large

# Set up your API key in youtube_summary_tool_copy.py
# Replace: api_key="YOUR_API_KEY_HERE"

# Launch the application
python app.py

1. Start the server: python app.py
2. Navigate to http://localhost:5000
3. Enter YouTube URL and wait 3-5 minutes for analysis
4. Explore results in Summary, Sentiment, Word Cloud, and Q&A tabs

• "What do people think about the music quality?"
• "Are there any complaints about the video?"
• "What are the most common suggestions?"
• "How many people mentioned the graphics?"

# Analyze a video
python youtube_summary_tool_copy.py "https://www.youtube.com/watch?v=VIDEO_ID"

# Ask specific questions
python youtube_summary_tool_copy.py "VIDEO_ID" --question "What are the main criticisms?"

# Custom sampling
python youtube_summary_tool_copy.py "VIDEO_ID" --question "Summarize reactions" --k 100

POST /api/analyze
Content-Type: application/json

{
  "youtube_url": "https://www.youtube.com/watch?v=VIDEO_ID"
}

POST /api/ask
Content-Type: application/json

{
  "question": "What do people think about this?",
  "k": 50
}

GET /api/status

When you submit a YouTube URL:

1. Comment Extraction (30-60s): Fetches all comments via YouTube API
2. Embedding Generation (1-2 min): Converts comments to vector representations
3. Database Storage (30s): Saves to ChromaDB for fast retrieval
4. Sentiment Analysis (1-2 min): AI analyzes emotional tone
5. Visualization (30s): Generates charts and word clouds
6. Summary Creation (1-2 min): LLM writes comprehensive summaries

When you ask a question:

1. Vector Search (milliseconds): Finds comments most similar to your question
2. Context Building (seconds): Assembles relevant comments
3. LLM Processing (30-60s): Llama 3.2 synthesizes an answer
4. Response delivered with context and insights

Web Interface ◄──► Flask Backend ◄──► YouTube API
     │                   │
     └───────────────────┼──────────────────┐
                         ▼                  ▼
                 Analysis Engine      ChromaDB
                         │           (Vector DB)
                    ┌────┼────┐
                    ▼    ▼    ▼
                Ollama HuggingFace NLTK
                (LLM) (Sentiment) (Text)

• Default: AmaanP314/youtube-xlm-roberta-base-sentiment-multilingual
• Alternative: VADER (available in youtube_summary_tool.py)

The system automatically calculates optimal sample sizes:

• <50 comments: 60-70% of total
• 50-200 comments: 30-40% of total
• 200-1000 comments: 20-30% of total
• 1000+ comments: 10-20% with caps at 600

Each analysis creates:

chroma/
└── VIDEO_ID/
    ├── sentiment_pie_chart.png
    ├── comment_wordcloud.png
    ├── overall_summary.txt
    ├── sentiment_summary.txt
    └── video_metadata.json

• Processing Time: Initial analysis takes 3-10 minutes depending on hardware
• API Limits: YouTube API has daily quotas (10,000 requests/day by default)
• Hardware Dependency: Performance heavily depends on GPU/CPU specifications
• Memory Usage: Large videos (5k+ comments) require significant RAM
• Language: Optimized for English, supports multilingual content
• Dependencies: Requires local Ollama installation and model downloads

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