AUTOMATICALLY TRAINING SPEECH SYNTHESIZERS

A method includes receiving, by a system, a voice recording associated with a user, transcribing, the voice recording into text that includes a group of words, and storing an association between a portion of each respective word and a corresponding portion of the voice recording. The corresponding portion of the voice recording is the portion of the voice recording from which the portion of the respective word was transcribed. The method may also include determining a modification to a speech synthesis voice associated with the user based at least in part on the association.