Is it just complete trial and error with modulation? Is there some sort of method to it?

(Im using FL Studio with Chipsounds rn just because I like the workflow)

if you're talking about stuff like this

then yes, more or less. by modulating pulse width you can mimic the formants of a voice.

i wouldn't say it's "trial and error" though, pulse width only modulates across a single axis so all there really is to do is decide if you want any given "vowel" to go from wide-to-narrow or narrow-to-wide.

its trial and error until u hav enough experience and it begins to be experience instead ;D