Facial recognition technology plays a crucial role in modern security systems, enabling efficient authentication and surveillance. This study presents a comparative evaluation of Haar Cascade and YOLOv8 frameworks for real-time face detection and recognition within Visual-bot (V-Bot), a system designed for headcount estimation and mood prediction across images, videos, and live streams. The experimental results demonstrate that YOLOv8 significantly outperforms Haar Cascade in accuracy, robustness, and real-time efficiency, achieving an average recognition accuracy of 92% under varied conditions—an improvement of over 30% compared to traditional approaches. The main contribution of this work lies in integrating advanced deep-learning models with modular system design to enhance security measures through multi-faceted facial analysis. Future applications of this technology span across automated surveillance, behavioural monitoring, and smart access control systems, particularly in high-security environments such as airports, banking institutions, and public spaces where real-time identity verification and behavioural insights are crucial.