For those who might find this effect useful in production there is a Steven Hawking voice box simulator on line. Here’s the link. https://voicemynah.com/
Having used similar text-to-speech generators before (especially the excellent VOKI.com), I’ve found that some “phonetic training”
is sometimes required – e.g. longer words can work “better” if split or as individual sylables etc e.g. I wouldn’t put in say “gameswizard”
instead do as the two words, also anything with vowels put in as the “sound” – especially dipthongs as in “Iain” needs the system to run it
as e.g. ee aainnn; some individual “voices” on such systems need more or less such “training than others.