Thanks to modern machine learning techniques, text-to-speech engines have made massive progress in recent years. It used to be incredibly easy to know that it was a computer reading a text, not a human being. But that changes quickly. Amazon's AWS cloud computing arm today released a series of new neuronal text-to-speech models and a new newsreader style designed to mimic the sound of news broadcasts.
"Voice quality This is certainly important, but more can be done to make a synthetic voice even more realistic and engaging," the company said today. "What about style? Surely human ears can tell the difference between a news program, a sports broadcast, a university class, and so on. In fact, most people use the right style of speech for the right context, and this certainly helps to convey their message.
The new news channel style is now available in two US voices (Joanna and Matthew), and Amazon is already working with USA Today and Canada's The Globe and Mail, among other things, to help them to pronounce their texts.
Listen for yourself:
Amazon Polly Newscaster Officially Called a New Service Research into Text-to-Speech, The AWS now also provides about its neural text-to-speech engine. This new engine, which is similar to similar neural engines like Google WaveNet and others, currently has eleven votes, three for British English and eight for American English.
In times of fake news, you can hear some more voices here.
In times of fake news -like robotic voices that sound like real news channels feel a bit problematic at first. Most of the time, however, it makes no difference whether a robot or a human reads the text. There are many good use cases for the voices. Based on the examples provided by AWS, you can listen to these voices much longer than the old ones before you want to cut off your ears.