Artificial intelligence already has a human voice
Artificial intelligence already has a human voice
Neural networks allow Google to interpret a text and apply similar intonations to a person
Siri, Alexa, Cortana; they are some of the virtual assistants of the American technological giants. One of these names was added a few years ago: the Google Assistant. Paa the search giant has put many resources to get unseat Apple as the king of the audience.
Through the Deepmind laboratory, responsible for the Artificial Intelligence (AI) of the company, Google has managed to develop a synthetic voice with a precision almost identical to that generated by a human being.
This technology is called Tacotron 2, whose operation is based on two neural networks capable of interpreting the text and applying the effects, pauses and intonations that are present in it.
From the American portal Quartz indicate that one of the novelties of this version is the ability to pronounce names and difficult words, as well as distinguish punctuation marks or if a part of the text requires to be pronounced with a different intonation.
This system allows you to transform the text into speech first, recreating the text in a spectrogram, that is, graphically forming the words in the form of waves so that the Google Wavenet software interprets them and then transforms them into a voice.
So far, Tacotron 2 is in the testing phase, available only in English and played by a female voice. Although, the company intends to train other versions in different languages and interpreted by a male voice.
Comments
Post a Comment