### Answer The pronunciation inconsistency you're experiencing can be addressed using Retell AI's pronunciation dictionary feature [(1) ]() . This allows you to specify exactly how names should be pronounced throughout the conversation . You can configure the pronunciation dictionary in your agent settings using the

pronunciation_dictionary

parameter . Here's how to set it up:

json
"pronunciation_dictionary": [
  {
    "word": "Jean",
    "alphabet": "ipa",
    "phoneme": "ʒɑ̃"
  }
]

The pronunciation dictionary accepts entries with three fields:

word

(the text to match) ,

alphabet

(set to "ipa" for International Phonetic Alphabet) , and

phoneme

(the pronunciation guide) . Why the pronunciation changed: The agent likely reverted to the default English pronunciation after the initial correct pronunciation because there was no persistent instruction telling the TTS system to maintain the French pronunciation. Without a pronunciation dictionary entry, the text-to-speech system defaults to standard language rules based on your agent's configured language setting . Additional considerations: You can also adjust the

voice_temperature

parameter to control voice stability . Lower values (closer to 0) produce more stable, consistent speech . Setting

voice_temperature

to a lower value like 0.5 or 0.7 might help maintain pronunciation consistency [(2) ]() . The

language

parameter in your agent configuration affects speech recognition optimization [(1) ]() . If you're working with French names frequently, you might consider setting the language to

fr-FR

or

fr-CA

for better French pronunciation handling .

​ To learn more, see the sources I considered: [[1] Update Voice Agent](https://docs.retellai.com/api-references/update-agent) [[2] Voice inconsistency - stuttering and filler sounds (aaaa, ehhh, hmmm)?](https://community.retellai.com/t/voice-inconsistency-stuttering-and-filler-sounds-aaaa-ehhh-hmmm/769) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Yes, you're correct—setting a pronunciation dictionary entry for "Jean" would affect all instances of that word throughout the conversation [(1) ]() . The pronunciation dictionary applies globally to the specified word, so if you set "Jean" to use French pronunciation (ʒɑ̃) , it would pronounce every "Jean" that way, including American English names . Potential workarounds: Based on the available sources, there isn't a built-in way to conditionally apply pronunciations based on context. However, you could consider these approaches: 1. Use dynamic variables to capture the name - Extract the caller's name during the conversation and potentially adjust pronunciation handling based on detected language or accent patterns [(2) ]() . 2. Leverage multilingual support - Retell AI supports 31+ languages with native-quality AI voices for global customer experiences [(3) ]() . If your agent handles multiple languages, the TTS system may automatically adapt pronunciation based on the detected language context [(2) ]() . 3. Community-reported approach - Some users have found that using phoneme-based pronunciation keywords can work for specific instances, though this may require testing with your specific TTS provider [(1) ]() . The normalize text for speech feature converts text to spoken form before TTS processing, which happens as a pre-processing step [(4) ]() . This normalization is language-dependent and uses the configured language code , but it do

esn't solve the context-specific pronunciation challenge you're facing. For a more sophisticated solution that handles context-aware pronunciation, you might need to implement custom logic in your agent's conversation flow to detect when French pronunciation is appropriate.

​ To learn more, see the sources I considered: [[1] pronunciations issues](https://discord.com/channels/1186424376300679209/1231651289201508502) [[2] How Multilingual Voice AI Drives Global Sales Growth | Retell AI | Retell AI](https://www.retellai.com/blog/multilingual-voice-ai-on-global-sales-growth) [[3] Top 5 Thoughtly Alternatives for AI Voice Agents in 2025 | Retell AI | Retell AI](https://www.retellai.com/blog/thoughtly-ai-voice-agent-alternatives) [[4] Normalize text for speech](https://docs.retellai.com/build/normalize-text) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Based on the available sources, you can extract and store information during a call using the Extract Dynamic Variable Node in Conversation Flow [(1) ]() . This allows you to capture the caller's name and other details in real-time [(1) ]() [(2) ]() . You can create variables with different types (Text, Number, Enum, or Boolean) [(1) ]() . For your use case, you'd extract the name as a Text variable . Here's what you can extract: - Variable Name – A short name to reference this variable - Description – A brief explanation of what this value should be - Variable Type – Choose from Text, Number, Enum, or Boolean The Extract Dynamic Variable Node lets agents "automatically save names, emails, budgets, and preferences" and "reference collected data anytime during the call" [(2) ]() [(3) ]() . However, the available sources don't show a method to dynamically adjust pronunciation handling mid-call based on extracted variables. The pronunciation dictionary is configured at the agent level and applies globally to specified words throughout the conversation. While you can extract the name during the call and store it as a dynamic variable [(4) ]() [(1) ]() , there's no documented way in the sources to use that extracted information to conditionally modify pronunciation settings in real-time duri

ng an active call.

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

​ To learn more, see the sources I considered: [[1] Extract Dynamic Variable Node](https://docs.retellai.com/build/conversation-flow/extract-dv-node) [[2] Major Platform Upgrades: Knowledge base, warm transfer, Voice & Analytics](https://www.retellai.com/changelog/major-platform-upgrades-knowledge-base-warm-transfer-voice-analytics) [[3] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) [[4] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) ​