Hey, the solution for this depends on two things:
1. If it's a one-time word where you require pronunciation to be corrected for this word only, then using prompting we can hack into it.
2. If it's in something which gets changed or it's a unique per call, then you have to use phonemics. As of now, I haven't tested them yet through 11 Labs.
Depending on which solution you require, I can further communicate to you. Please let me know if this is the case - if it's a case 1 or case 2. Apologies for the delay. I tried a few approaches discussed with the team, and we finally came to a conclusion that we might have to give support for phonemics, which could solve the issue for you and for others who require correct pronunciation.