Google said it built on voice datasets for the tool. Using a robotic-sounding voice without any natural cadence wouldn’t be conducive to listening to long-form content, and many listeners would likely find it distracting. For this tool, Google focused on making the voices sound more “natural” and “expressive” while reading content aloud.

Advertisement
Advertisement

David Kadouch, Google’s product lead for research and machine intelligence, said in a video that the company’s goal is to create a “smooth, fast, and natural-sounding experience” for the Assistant reader. The team considers three primary areas—stress, intonation, and rhythm—to make the voice sound more human. And while a short demo clip indicates that there’s still a lot of room for improvement, it does—to its credit—sound smooth to the extent that you could probably listen to it for the entirety of the length of an article without too much distraction.

Google says the tool will be available later this year on Android phones versions 5 and above.