This automator action converts the selected text into speech (much faster than realtime, but not exactly instantly) and creates an AIFF file of it. I’m pretty sure it’s just doing this terminal command:

say "sample text that goes on and on" -o audio.aiff

I say “pretty sure” because this is total speculation and I can’t find any evidence to support it. It’s just that they seem to do the exact same thing in about the same amount of time. And Apple isn’t usually big on having multiple different ways to do the same exact thing.

How is it doing this? Where does this automator action call this command, if that’s what it’s doing?

There are no executables. Many other automator actions have executables. Why not this one? There are several property lists, but according to Spotlight none of them have the word “say”. There is a binary nib for each of 19 different languages, though. Could they have embedded a command in the nib? Is that possible? If so, wouldn’t it be wasteful to have the same exact command in 19 different places?

This is the stuff I try to figure out at 11:30 at night.