Speech changes to a more robotic feel if I use certain phonemes

Google Cloud Build with AI AI APIs
1

I am using phonemes for some speaks, to make them sound more natural, but some phonemes change the rest of the sentence feel. I have made a demo here.

Using these settings:

  • English
  • EN-GB-Wavenet-F
  • speed 0.95
  • pitch 0

In the speak below I have two identical speaks, where one word is replaced with a phoneme. Notice how the end of the sentence changes to a more robotic feel in the first.

How do I avoid this?

In this training, you will learn more about how you sell placeholder as a solution for companies that want to minimize out-of-pocket spending and have better control of company spending by employees.

In this training, you will learn more about how you sell card as a solution for companies that want to minimize out-of-pocket spending and have better control of company spending by employees.

2

Hi! I see there is an issue already open regarding this. I’ve tested your SSML and I understand what the issue is and it seems like unintended behavior. I tried to workaround it but I didn’t find a way to do so. Any updates regarding your issue will be on the issue tracker

3

Yes. I created the issue. Thx for the reply