Method and system for personalised voice synthesis

Disclosed is a method for personalised voice synthesis. The method comprises obtaining an audio of a person speaking the words of a natural language text content. The audio is processed to identify, for each speech unit (which may comprise a phoneme) of the given natural language, at least one audio frame range where the speech unit is occurring in the audio. For a given word to be synthesised in the person's voice a sequence of speech units is first identified occurring in the given word. Then, individual audio frames of the speech units occurring in the given word are extracted from the obtained audio. The extracted audio frames of the speech units occurring in the given word as per the identified sequence are then used to produce a voice-synthesised audio of the given word in the person's voice and accent. The audio frame range may consist of speech units occurring at the beginning, middle or end of a given word.