High-quality and flexible speech synthesis with segment selection and voice conversion