Method and system of producing natural-sounding recitation of story in person's voice and accent

摘要

Disclosed is a method of producing a personalised natural-sounding recitation of a story in a persons voice and accent. The method comprises obtaining an audio of the person reading aloud words of a text content, which is different from the story text. From this audio, individual audio frames containing identified speech units of the given natural language are extracted. The speech units in the extracted audio frames are concatenated to produce a voice-synthesized audio of the given word and/or sentence in the persons voice and accent. A predefined time gap profile is used to insert different time gaps between different pairs of words occurring substantially consecutively in the story text, when producing the recitation of the story in the persons voice and accent. The amplitude of given words may be adjusted based on determining a predefined special symbol within the text.