The present disclosure describes techniques for annotating a video. The techniques comprises receiving an operation of creating a bullet screen of annotating a target object in a first frame; determining a display position of the bullet screen in the first frame; displaying the bullet screen at the display position in the first frame; determining an offset of the display position of the bullet screen relative to a position of the target object in the first frame; and displaying the bullet screen at a target display position in a second frame of the video, wherein the target display position in the second frame is determined based at least in part on the offset of the display position of the bullet screen relative to the position of the target object in the first frame, the second frame is played subsequent to the first frame, and the second frame comprises the target object.