Social media platforms generate vast amounts of opinion-rich data that reflect public emotions and attitudes. The project explores sentiment analysis using deep learning models to classify emotions in user-generated text. The system achieves high accuracy in identifying emotional polarity, helping interpret public opinion effectively. The project uses YouTube Data API, social media posts to extract comments from video links posted on social media. Comments are cleaned and passed through a multilingual sentiment analysis model based on BERT to classify emotions. The model supports both English and regional language inputs, including code-mixed text. Visual results are displayed as pie charts along with a comment-wise sentiment summary. The tool enables real-time sentiment extraction from video discussions using deep learning.