The voice assistants on our smartphones wouldn’t be possible without advanced speech-synthesis tech, allowing the software behind these systems to turn responses into natural-sounding spoken replies. And while today’s Cortanas, Siris, and all their peers do an admirable job, there’s always the desire to add a little more of a human-sounding element to the process, keeping them from sounding like robots. Google’s been working on just those kind of improvements, and in a new video shows us how it’s bringing a new and improved voice to Google Now and its voice-driven search.
We get to see what Google’s been up to with the help of Nat and Lo, the same pair of Googlers who have helped us out in the past with things like some early guesses at Android M launch names. Here, they learn about how speech synthesis works, and how even when it’s capable of generating perfectly understandable output, there are still ways to minimize that and create a more natural-sounding tone.
That includes paying closer attention to intonation and other minor difference in how we say the same things in different situations – how questions sound different from statements. That involves using both speech generated from combining lots of parts of shorter recordings, as well as longer segments custom-recorded in response to likely queries.
You can check out the whole process below, or turn to your phone to hear this subtly enhanced voice action for yourself.