Voice Morphing Based on Formant Substitution and Phoneme Interpolation

Voice cloning is a major topic of speech sound processing and artificial intelligence. Its possible applications include customization of generated voice in robots, text-tospeech, it can also be used for making lines of speech of a given voice actor when he/she is unavailable. In this paper we describe a method to change the perceived speaker into a desired one by substituting the formants, interpolating between known phonemes when the measured formants don’t exactly coincide with them. The method enables voice cloning with a short sample of target voice.