Best way to generate clear chemistry narration with pronunciation controls?

Hi, I’m exploring Genny for adding optional spoken explanations to Chemistry AI (https://chemistryai.chat/), where answers can contain formulas such as H2SO4, equilibrium arrows, oxidation states, and IUPAC names.

Is there a recommended Genny API workflow for: (1) defining pronunciation overrides for chemical notation and compound names, (2) inserting short pauses between equation steps without making the speech sound unnatural, and (3) keeping generated audio timing reasonably stable when only one line of an explanation changes?

I’d also be interested in whether pronunciation dictionaries can be reused across requests or voices, rather than sending all replacements with every generation call.