We present PixTrack, a vision based object pose tracking framework using novel view synthesis and deep feature-metric alignment. Our evaluations demonstrate that our method produces highly accurate, robust, and jitter-free 6DoF pose estimates of objects in RGB images without the need of any data annotation or trajectory smoothing. Our method is also computationally efficient making it easy to have multi-object tracking with no alteration to our method and just using CPU multiprocessing.