How do I get Actions on Google to speak using Google Home's normal voice?

Question

I'm building an app using Actions on Google using API.AI that allows users to speak to my app through Google Home. API.AI allows me to choose between two male and two female voices, however, all four of these voices are not the normal Google Home voices.

Is there any way to get Google Home to keep its normal voice rather than these four robotic alternatives? It's strange that Google Home suddenly switches voices.

Araby

Answer 1

There is no way to use the "default" voice, and the change in voices as well as the starting and ending chirp are both a security and a UI design principle.

The change in voice lets the user know they are no longer talking to the Assistant, but are talking to another entity (your Action). This is a rough parallel to a GUI that makes it clear you're running a user program and not one of the system programs. It also cues you that the usual system commands (adjusting volume, for example) won't work.

Take the opportunity to use the voice options to help set the environment for your Action and the tone that you'll be using to talk to the user.

Dulcia

Thanks for the info! I totally get how a change in voice would give the user a cue that it's a separate system they're talking to, but it's so disappointing that the available voices are so terrible in comparison to Google's amazing voice. It's just strange that the voice just suddenly degrades into what sounds like a robot in comparison. – Araby

I agree, the four voices are much worse. – Keijo

I fully agree with @DanLeveille! I possibly disagree with the reason why using on an third party action a voice different from the default Google Assistant voice, is a good choice. It could be an option. – Unpredictable

Answer 2

Wavenet voices should sound more natural than the old voices. You can hear the new voices here and they say that they should be rolling out in the next few weeks.

Migrate to high-quality voices Starting March 4, 2019 the Actions on Google platform will support higher quality Wavenet voices for a subset of languages and locales. These voices are rated with a mean opinion score (MOS) of greater than 3.6, thus providing listeners with higher voice fidelity. We will automatically apply the new voices if there is a new voice for the voice option used in your project.

Surbased