Google's Artificial Intelligence speaks like a human

Google’s Artificial Intelligence speaks like a human

January 2, 2018

Tacotron 2 is the new neural network architecture developed by Google that takes synthetic speech generation to the next level. Tacotron 2 combines functions from previous Google projects, such as WaveNet and Tacotron, whose objective is to train machines to speak like humans.

Image by Sergey Nivens via Shutterstock

The idea with Tacotron 2 is to achieve that the synthesized voice can, for example, produce a fluid and natural speech, from text, without having to train with a large amount of metadata about language and grammar, so that the dynamics work correctly.

The Google team has shared a series of audios on GitHub, where it challenges us to identify which is the synthesized voice, and the human voice, since the results they are achieving with Tacrotron are amazing.

Keep reading: The synthetic voice created by Google’s AI sounds almost human

[+] Videos de nuestro canal de YouTube