Microsoft has revealed its newest AI model called VALL-E, which they say can synthesise 'high quality personalised speech' using just three seconds of sample speech.