yes, i'm aware that this tool i wrote in js requires js lol. the audio here is being generated by an actively running vocal synth ported from a UI-only project, and hacked into an actively running game engine. it's not just generating a buffer for a full phrase and queuing it for playback after the fact. an audio export would require rewriting both of these such that you could deterministically generate the audio for the phrase as a single function call like that (which is possible, but like i said, not trivial).
it's all open source though so feel free to prove me wrong if you think i'm overestimating!