SHIRO is a set of tools based on HSMM (Hidden Semi-Markov Model), for aligning phoneme transcr iption with speech recordings, as well as training phoneme-to-speech alignment models.