Have you tried a different voice model? Just to check if it's specific to that one.