Re:ATT services down due to bombing in Nashville
by vsareto on Saturday December 26, 2020 @09:06AM
(#25543515)
Looks like it may be Microsoft Anna text-to-speech. The poster added reverb and echo and it was very, very close to the youtube security video that captured the warnings. The device that recorded that may have also skipped on its own recording and may not be perfect.
https://www.youtube.com/watch?v=8B--BNdnuvA
The original video:
Re:ATT services down due to bombing in Nashville
by doctorshady on Saturday December 26, 2020 @01:11PM
(#25545181)
This is probably a much more likely explanation; we'll see when we get bodycam footage or anything else if there's a better recording of what was being played. But for whatever it's worth, most TTS engines are modeled around well known voiceover people.
It's still possible that since the engine didn't reproduce a 1:1 copy and there's no discernable (at least to me) artifacts in the one copy of sound footage we have, that was the real person, and she just announced it slightly differently.
For the purposes of arguing, let's say someone got ahold of some municipal evacuation recordings and chopped them up for this. It's not likely they had enough phrases to warn people a bomb would be going off in specifically fifteen minutes.
EDIT: For anyone curious about other instances of bad edits, listen to the word "this"; there's another word before that, and it was cut off: "thi-this area must be evacuated now". Also, "if you can hear this message, <unnatural pause> evacuate now." It's not likely any professional voiceover person would announce this way, nor would anyone who's paid attention to basic edits before consider this passable.
If this is all true, I guess the next question to ask is who would have access to those announcements, and who would think to use them? Most people would probably just use some text-to-speech thing unless it was easy/they had good reason. I'm probably overthinking this, though.