"We've used 5% of our defense budget to arm Ukraine. And with it, they've destroyed 50% of Putin's army." - Ad targeted to Republicans to support Ukraine
Thinking out loud, I wonder if it's better to aim computer generated voices to be understandable and pronounce words correctly, but place them firmly on the lower side of the Uncanny Valley. In other words, let them be noticeably different, but otherwise comprehensible.
For ads and movies and such, obviously those companies want the most realistic voices they can. But for voice assistants, maybe it's better if we don't.
Since i work in media i can tell you with 99.9% certainty that's not an AI-generated voice. It's just cut a bit too tightly and the super low quality of the audio track doesn't help either.