Is it painless to make this a drop-in replacement for the Linux utility "spd-say"? Or will it always take ages to go from inputting text to hearing a voice?

This is pretty amazing. Now guys can be vthots on YouTube.

I only know that the site might be buggy because of traffic, but they should be fixed in the next few days. As for the spd-say drop-in replacement, that would possibly using cURL, PhantomJS, whatever, to simulate requests a web browser would make, and I don't know much about both. Maybe leave that to someone else... or simply use a web browser for this (it looks like it requires enabling some 3rd-party requests in uMatrix - also, does it work in Pale Moon?)

I ended up downloading a few voice samples for the things I use spd-say in my scripts and pick one at random each time. Now I'll be getting a more pleasant voice experience when things go wrong. (I went with spd-say because it was too easy to overlook notifications.)

interesting, is it open source?

I don't think so, but I think you can use the voices for anything as long as you give credit