Karuna Sree, K.Suchithra, D.Srilekha,S G.Gowthami, S.Sindhu, A.Vijaya Laxmi 02 06 The exponential rise of shortform videos on platforms like YouTube Shorts TikTok and Instagram Reels has led to increased concern over the spread of hate speech often hidden in the audio tracks of these videos Traditional textbased detection methods fail to capture the nuances of spoken content This project proposes an audiobased hate speech classification system using deep learning techniques that analyze speech patterns tone and content from video audio By leveraging audio preprocessing feature extraction eg MFCCs and classification models such as LSTM and CNN the system can identify hate speech even in disguised or nuanced speech forms This model offers a scalable and automated solution to moderate content effectively on social media platforms Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026