MOTRv2 is a computer vision system for multi-object tracking in video. It bootstraps an end-to-end tracker using a pretrained object detector (YOLOX) to generate detection proposals as anchors, improving upon prior end-to-end approaches. The system uses transformer-based architecture and achieves state-of-the-art performance on large-scale benchmarks like DanceTrack (73.4% HOTA) and BDD100K.