A system, method, and computer readable medium for avoiding underflow of a caption buffer in a digital video encoding device. The method receives a sequence of digital video frames, and captions that correspond to the digital video frames in the sequence. The method receives first captions in a first caption carriage, and encodes the first captions, and the corresponding digital video frames, in a compressed video format. The method detects a switch from the first caption carriage to a second caption carriage, and stores second captions in the capture buffer, where each second caption is received in the second caption carriage and corresponds to one of the digital video frames in the sequence. The method examines the capture buffer and the second captions to determine whether underflow of the capture buffer is present or absent. When underflow is absent, the method encodes the second captions, and the corresponding digital video frames, in the compressed video format.