I am trying to launch a new inbound routing flow for our Australian contact centre and the text-to-speech engine is driving me absolutely crazy. We are using the native Genesys TTS voice ‘Olivia’. It sounds okay for basic greetings, but whenever it tries to pronounce local Australian town names like ‘Woolloomooloo’ or ‘Wagga Wagga’, it completely mangles them. Our customers are complaining that it sounds unprofessional. I tried spelling them phonetically in the Architect ‘Play Audio’ action, but it still sounds wrong. How do we force the TTS engine to actually pronounce these regional locations correctly?
Hello! I am so happy to help with this! When integrating complex systems, ensuring accurate localized pronunciation is absolutely critical for the customer experience! You cannot simply rely on phonetic spelling within the basic text box. You must utilize Speech Synthesis Markup Language (SSML)! Instead of just typing the word, you need to format the text using the <phoneme> tag. For example, you would write <phoneme alphabet="ipa" ph="ˈwʊləməlu">Woolloomooloo</phoneme>.
This bypasses the standard TTS dictionary and forces the voice engine to synthesize the exact phonetic sounds! It requires learning the International Phonetic Alphabet (IPA), but it gives you flawless control over regional dialects!
While embedding SSML tags directly into individual Architect ‘Play Audio’ actions is technically viable for isolated instances, it represents a significant operational risk for long-term maintenance. If a town name is utilized across twenty different call flows, updating the SSML in twenty different locations is highly inefficient. The enterprise standard is to utilize the centralized ‘Pronunciation Lexicons’ feature within the Genesys Cloud Admin interface.
You construct a global XML lexicon dictionary mapping the complex regional terms to their IPA phonemes. You then assign this lexicon globally to your TTS engine, ensuring standardized pronunciation across the entire organizational architecture.