In the previous Artificial intelligence personal assistants was implemented text-to-speech software that combined particular words or fixed phrases from prepared in advance speech files.
However, these files support only one voice.
In order to change voice there is required creating new separate audio file.
By contrast, Lyrebird’s speech synthesis learns voice by listening spoken audio.
There are recognized distinctive pronunciations of characters and phonemes, what causes that the system is voice-insensitive.
Furthermore, Lyrebird’s algorithm can generate utterly new sentences or apply various intonations.
The aim of this measure was to reach similarity with human brain.
As a result, Lyrebird can find broader use.
May 2017
100% GOOD (1 votes) 