In this paper, we present ELT-Easy Label Trainer (ELT), a comprehensive application designed to streamline the process of labeling and training object detection models while tracking objects for analysis. Leveraging advanced models such as YOLOv8, YOLOv9, and YOLO-World, ELT provides an intuitive interface built with Gradio for users to load video streams and perform object detection or segmentation. YOLO-World, known for its open-vocabulary detection using descriptive text prompts, is enhanced within ELT by custom prompts and additional parameters to improve detection confidence. Despite its robustness, YOLO-World’s performance decreases for objects outside the COCO128 dataset. To address this, ELT employs pseudo-labeling to capture low-confidence labels and retrains models with YOLOv8 or YOLOv9, enhancing detection accuracy. ELT also automates object tracking and labeling using the Segment Anything Model (SAM) and Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model (XMem), inspired by Track Anything, to facilitate precise and efficient labeling. We demonstrate the effectiveness of ELT on GMOT-40: A Benchmark for Generic Multiple Object Tracking datasets, showing robust object detection in real-time. The results indicate a significant reduction in manual labeling effort and enhanced model performance, proving the viability of ELT in diverse practical applications.