To understand much better what is and how it works, in broad strokes, a system of synthetic voice I want to refer to a clear example that surely we have all come across at some time, specifically I am talking about those videos present on YouTube as well as other internet services where the narrator speaks through a computer generated voice. Perhaps the best known and most widely used reading software is crazy Although today the truth is that these systems have evolved a lot, we have the proof in Cortana o Crab.
Today the latest and sophisticated speech synthesis program presented by Google, a software known under the name of Waynet and that has been created by the engineers belonging to the department DeepMind, an artificial intelligence company that was acquired by Google in 2014. WayNet is a speech synthesis software based on complex artificial intelligence algorithms which functions as a complex neural system.
WaveNet, a revolutionary voice synthesizer that will surprise you
Among the novelties that WayNet presents, it should be noted that, although until now the main method used was the TTS, text to speech, where different recorded speech fragments were combined to build words and sentences, or known as Parametric TTS, a method that sends the text to a speech coder whose results are even less natural than the previous one, we now find that WayNet, instead of just combining and playing audio, integrates a complex artificial intelligence system that is capable of learning and adapting to the context.
This new system is capable of performing 16.000 samples per second allowing you to even generate your own audio sequences without human intervention. On the other hand, it is worth mentioning that the engineers responsible for its development have introduced a system capable of resorting to statistics to predict what it will have to say later and thus ensure that the system offers results much more quickly and fluidly. If you are interested in WayNet, tell you that on its website you can listen to various samples in English and Mandarin Chinese.
Further information: DeepMind