Google Artificial Intelligence (AI) technology evolved, the company announced that the robot can make the tone no longer blunt, sounds and human beings difficult to distinguish between true and false.
PhoneArena, Quartz, The Daily Mail reported that the Google blog has claimed that it has developed the second-generation text-to-speech system “Tacotron 2” that makes the machine speak no strange, unnatural sounds that sound almost identical to humans. The machine will determine the correct pronunciation based on semantics, the article spelling mistakes can be easily read out, and even around the password are hard to fail.
Google officials said Tacotron 2 is considered a professional like the outside world. The average Opinion Score for the public at Tacotron 2 was 4.53 points, only slightly below the 4.58 points recorded by professionals.
The Tacotron 2 includes two deep neural networks, the first of which converts text into a spectrogram, images the audio, then enters the spectrogram into the WaveNet system and the machine emits sound as shown (Tacotron 2 and human voice control point).
However, Google said that Tacotron 2 still needs improvement in many areas, including the inability to read some of the dubs and the inability to synthesize audio on the fly; the machine’s voice has no emotions and does not convey the feeling of happiness or sadness.
Despite this, there is still plenty of room for this technology, with Google Voice Assistant currently using WaveNet technology to make Google Assistant even more viable with Tacotron 2.
share by: YUYI Global Technology