Biometrics deals with the problem of uniquely identifying individuals based on physiological or behavioral attributes. Physiological biometrics involve measurement from body parts such as the fingerprint, face, iris, etc., whereas behavioral biometrics exploits cues such as gait, voice, expressions, etc. In this chapter, we discuss video-based biometrics involving faces and gait. We discuss spatiotemporal models appropriate for each task, followed by design of metrics for classification. We discuss how careful modeling of the variations of appearance and motion leads to improved biometric systems.