Compression algorithms for speech, audio, still images, and video are quite complicated and, more importantly, nearly always lossy. Thus, samples often change dramatically once they’re decompressed.