This study presents a real-time object recognition system specifically designed to support visually impaired individuals by improving their independence and environmental awareness. Unlike general-purpose object detection systems, the proposed approach is designed with accessibility as a primary focus. It emphasizes intuitive interaction and clear auditory feedback to better support the needs of visually impaired users. The system combines computer vision, natural language generation, and speech synthesis within a mobile application. The system then provides immediate audio descriptions of nearby objects and their positions in a clear and understandable way. The architecture consists of a React Native-based mobile frontend and a Python-based backend deployed on cloud infrastructure. A pretrained YOLOv8 object detection model is utilized to identify objects in real time. Spatial analysis is performed to determine the relative position of objects within the scene. Experimental results demonstrate that the application achieves reliable object detection with low latency. Furthermore, the user-centered design, including bilingual support and touch-based interaction, enhances usability and accessibility. Overall, the proposed solution contributes to the field of assistive technologies by offering an efficient, cost-effective, and portable tool. It improves environmental awareness and supports independent mobility for visually impaired individuals.