Robot Not a Dev? Pre-order Now
Not a Dev?

Synthesized Speech


#1

This story is about Erica, a humanoid robot that will soon become a TV news anchor in Japan. She has been described as having “one of the most advanced artificial speech systems in the world.”

Between Erica, Sophia, Siri, Google Home, Alexa, Jibo, Cortana, and any other synthesized speech system you’ve heard, in your opinion, which system has the best artificial speech? What is your opinion of the output of current speech synthesis algorithms? Which one sounds the best? In your opinion, do any of them pass the Turing test (Turing test - Wikipedia)?


#2

I haven’t heard them all, but between Alexa, Siri, Google Home and Jibo I would go with Google’s Home. The range of option seems and response variety seems the highest in my opinion. None comes anywhere close to passing the Turing test.


#3

Yeah, the tech behind the Google Home voice appears to be light years ahead of everyone else. They departed from the traditional concatenative approach a few years ago (piecing together pre-recorded tones from a database) when they introduced WaveNet - which uses a convolutional neural network to generate the actual sound waves on the fly. The model can then be trained on different sets of data with variations in accents, languages, etc…

This post shows how the Google voice now is very similar to a humans:

Compare that to Amazon’s Polly service - which still sounds like a robot from the 1990s!


#4

Nobody mentions Cortana :stuck_out_tongue:
Personally I thought Cortana sounded the most natural last time I did a comparison, though Google has made huge improvements recently.


#5

As am Samsung 'alum" I am obligated to be the only person to mention Bixbsy :wink: