This paper presents a way to generate a stained glass animation from a given video input. We first obtain low-frequency components from input video frames to get rid of textures which cause over-segmentation in image segmentation. Then we segment input video volume by employing mean-shift video segmentation. The segmented regions are too large to be architecturally stable, subdivision is required. To sub-divide regions temporally coherent, we obtain the panoramic image from the segmented regions, and sub-divide them by using weighted Voronoi diagram. To render these sub-divided regions as stained glass pieces, we find the best match glass piece in the real stained glass piece image database and transfer its color to that region. Finally, we generate lead came at the boundary of regions, which results in a temporally coherent stained glass animation.