"Say" was included with versions all versions of Workbench from 1.2 to 2.1. You should find it in the "utilities" drawer on the "extras" disk.
It is a simple UI for the translator.library and narrator.device, which were included with later versions of Workbench without the Say program.
It is actually DEVS:narrator.device which generates the synthesis, and LIBS:translator.library which parses text strings into phoenomes for the narrator.device.
Some versions of WB have the entry "Speak:" in their DEVS:mountlist file, so the Amiga will generate speach synthesis for any text string sent to that device.
So you do not require the "Say" program to use these features.
In summary, You don't require any extra HW or SW. All you need is a standard workbench install to be able to use the Amiga's speech synthesis.
Also, any game/application which requires these system files will probably include them on the disk.
Finally. That you even thought it was possible for an ST game to have speech makes me chuckle.