This study introduces an AI-powered assistive system that combines audio feedback and object detection for visually impaired people. Three models were thoroughly evaluated, YOLOv5 (IoU: 0.142), YOLOv8 (IoU: 0.53), Faster R-CNN (IoU: 0.95), and SAM. SAM performed better than the other models. It achieved 0.9947 average IoU. A dynamic audio system that creates spatial descriptions (e.g., “a medium closed door located at the bottom-left”) based on item position is integrated with SAM’s excellent segmentation capabilities. SAM outperforms the other architectures in accurate object localization, as shown by experimental data. The next step is to expand the range of objects it recognizes and improve its ability to work in busy, changing environments like crowded streets. This research bridges advanced AI with practical needs, offering a solution that could improve day-to-day independence for visually impaired users.