I think Google could do a better job if they really wanted to. It's just a question of where the resources are being prioritized. There are also other text-to-speech solutions out there, and DuoLingo could choose to use them. Or they could choose to use an actual voice. Again it's a question of priorities and resources.
I personally would love to have a choice of actual voices. I'd be willing to volunteer my time to record my voice doing English, if it would help in any of the English courses, because I love DuoLingo and feel grateful for it and would want to give back.
Also, look at the other languages. It's proof that they can do better. The Spanish and German courses have excellent pronunciation with only sporadic glitches. Even Russian, which is in Beta, seems to do better than this.