Any one here got tips on collecting email addresses and making sure the AI gets it correct especially in foreign language like Spanish? I'm using Deepgram Flux Multi

I've had good results by asking users to spell their email character-by-character, then repeating it back for confirmation before saving. For Spanish, it's also helpful to normalize common pronunciations (e.g., "guion bajo" = underscore, "punto" = dot) and use a validation step that reads the email back slowly. If you'd like, I can help optimize your Deepgram Flux prompt and email capture flow to improve accuracy. @Joseph_

Use a spelling/confirmation step have the AI read the email back letter by letter and ask the caller to confirm. Also add a validation prompt that flags anything that doesn't match standard email format before saving. For Spanish specifically, Deepgram Flux Multi handles it well but watch out for letters like @ being heard as arroba make sure your prompt accounts for that. I've worked on multilingual voice agents and this is a solvable problem. Happy to help if you want a second pair of eyes on your setup. @Joseph_