The detection of bias in media content, particularly podcasts, is crucial as digital platforms grow in influence. This research introduces a bias detection system that combines NLP, machine learning, and transformer-based models like Sentence-BERT. Using a custom-labeled dataset of over 4000 sentences across Political, Economic, Cultural, and Racial biases, generated via prompt engineering, the system identifies nuanced biases. Advanced preprocessing, contextual embeddings, and parallel processing ensure scalability. Also, real-time summarization of the content using Hugging Face’s BART improves the content’s usability. Practical deployment is made possible by a Streamlit-based web application that provides bias scores and ranks for more specific information. Analysis of outcomes indicates high accuracy levels, which indicates that the system can be useful for automated media bias assessment.