I'm pretty sure they didn't get any real vocal training, like watching levels, listening to yourself, etc.
"Here: Record this for the radio feed." But hey, they're paid to forecast WX.
Currently readability of the synthesized voices: Pretty darn good.
Local color and real human voice of the old way: I still miss it (we had some really nice voices in the Raleigh NC office, years ago, just untrained), a little bit.