<p>Nowadays, the social media is becoming main way for people to communicate and express themselves. In this environment, understanding the emotional content conveyed in multimedia content has become very important. The ability to analyze emotions in videos shared on social media platforms can provide important understanding about user sentiment, engagement and preferences. This knowledge can be used for personalized content delivery, targeted advertising as well as sentiment analysis among other things. The significance of emotion recognition in social media videos - including speech and facial image analysis - is very high because recently the social media platforms have become main channels for communication and expression. Therefore understanding emotional context carried by such multimedia content is crucial now more than ever before. Hence. this work developed a unique deep learning model, EmoNet for accurate emotion recognition in social media videos. In this framework, the Convoluted Fusion Network (ConFusionNet) for feature extraction and Attentive Multimodal Fused Transformer (AMFT) for emotion categorization are implemented to ensure an effective recognition. Moreover, the EmoNet provides unmatched accuracy and dependability in decoding emotions from multimedia content by utilizing AMFT’s multimodal fusion capabilities and ConFusionNet’s capacity to catch intricate variations in emotional expression. Based on outcomes, it is noted that the proposed EmoNet performs better than the recent state of the art methods, with an impressive 99% accuracy and 0.12% loss. These results highlight EmoNet’s promise as a cutting-edge approach to emotion recognition tasks, with wide-ranging uses in content analysis, affective computing, and human-computer interaction.</p>