Social media platforms, particularly Twit- ter, generate massive volumes of real-time textual data that reflect public opinion, emotional tendencies, and emerging societal trends. Analyzing this stream of infor- mation manually is infeasible due to its speed, scale, and linguistic complexity. This paper presents an enhanced real-time Twitter Sentiment Analysis system that inte- grates Natural Language Processing (NLP) models with live tweet fetching using the Twitter API v2. The pro- posed system employs a hybrid pipeline consisting of the VADER rule-based sentiment analyser for fast polarity detection and Transformer-based models for deeper con- textual sentiment classification. Additionally, the system incorporates optional modules for emotion recognition and toxicity analysis, enabling multi-dimensional inter- pretation of user-generated content. A Streamlit-based in- teractive interface allows users to fetch tweets in real time, analyze sentiment distributions, examine top key- words, and download processed outputs. The architec- ture is designed for scalability, efficiency, and accessibil- ity, offering a low-cost yet powerful solution for social sentiment monitoring and data-driven decision-making. Experimental evaluations demonstrate that the model combination improves interpretability and accuracy while maintaining responsiveness suitable for real-time applications.